Logic Apps Agentic Workflows with SAP - Part 2: AI Agents
Part 2 focuses on the AI-shaped portion of the destination workflows: how the Logic Apps Agent is configured, how it pulls business rules from SharePoint, and how its outputs are converted into concrete workflow artifacts. In Destination workflow #1, the agent produces three structured outputs—an HTML validation summary, a CSV list of InvalidOrderIds , and an Invalid CSV payload—which drive (1) a verification email, (2) an optional RFC call to persist failed rows as IDocs, and (3) a filtered dataset used for the separate analysis step that returns only analysis (or errors) back to SAP. In Destination workflow #2, the same approach is applied to inbound IDocs: the workflow reconstructs CSV from the custom segment, runs AI validation against the same SharePoint rules, and safely appends results to an append blob using a lease-based write pattern for concurrency. 1. Introduction In Part 1, the goal was to make the integration deterministic: stable payload shapes, stable response shapes, and predictable error propagation across SAP and Logic Apps. Concretely, Part 1 established: how SAP reaches Logic Apps (Gateway/Program ID plumbing) the RFC contracts ( IT_CSV , response envelope, RETURN / MESSAGE , EXCEPTIONMSG ) how the source workflow interprets RFC responses (success vs error) how invalid rows can be persisted into SAP as custom IDocs ( Z_CREATE_ONLINEORDER_IDOC ) and how the second destination workflow receives those IDocs asynchronously With that foundation in place, Part 2 narrows in on the part that is not just plumbing: the agent loop, the tool boundaries, and the output schemas that make AI results usable inside a workflow rather than “generated text you still need to interpret.” The diagram below highlights the portion of the destination workflow where AI is doing real work. The red-circled section is the validation agent loop (rules in, structured validation outputs out), which then fans out into operational actions like email notification, optional IDoc persistence, and filtering for the analysis step. What matters here is the shape of the agent outputs and how they are consumed by the rest of the workflow. The agent is not treated as a black box; it is forced to emit typed, workflow-friendly artifacts (summary + invalid IDs + filtered CSV). Those artifacts are then used deterministically: invalid rows are reported (and optionally persisted as IDocs), while valid rows flow into the analysis stage and ultimately back to SAP. What this post covers In this post, I focus on five practical topics: Agent loop design in Logic Apps: tools, message design, and output schemas that make the agent’s results deterministic enough to automate. External rule retrieval: pulling validation rules from SharePoint and applying them consistently to incoming payloads. Structured validation outputs → workflow actions: producing InvalidOrderIds and a filtered CSV payload that directly drive notifications and SAP remediation. Two-model pattern: a specialized model for validation (agent) and a separate model call for analysis, with a clean handoff between the two. Output shaping for consumption: converting AI output into HTML for email and into the SAP response envelope (analysis/errors only). (Everything else—SAP plumbing, RFC wiring, and response/exception patterns—was covered in Part 1 and is assumed here.) Next, I’ll break down the agent loop itself—the tool sequence, the required output fields, and the exact points where the workflow turns AI output into variables, emails, and SAP actions. Huge thanks to KentWeareMSFT for helping me understand agent loops and design the validation agent structure. And thanks to everyone in 🤖 Agent Loop Demos 🤖 | Microsoft Community Hub for making such great material available. Note: For the full set of assets used here, see the companion GitHub repository (workflows, schemas, SAP ABAP code, and sample files). 2. Validation Agent Loop In this solution, the Data Validation Agent runs inside the destination workflow after the inbound SAP payload has been normalized into a single CSV string. The agent is invoked as a single Logic Apps Agent action, configured with an Azure OpenAI deployment and a short set of instructions. Its inputs are deliberately simple at this stage: the CSV payload (the dataset to validate), and the ValidationRules reference (where the rule document lives), shown in the instructions as a parameter token (ValidationRules is a logic app parameter). The figure below shows the validation agent configuration used in the destination workflow. The top half is the Agent action configuration (model + instructions), and the bottom half shows the toolset that the agent is allowed to use. The key design choice is that the agent is not “free-form chat”: it’s constrained by a small number of tools and a workflow-friendly output contract. What matters most in this configuration is the separation between instructions and tools. The instructions tell the agent what to do (“follow business process steps 1–3”), while the tools define how the agent can interact with external systems and workflow state. This keeps the agent modular: you can change rules in SharePoint or refine summarization expectations without rewriting the overall SAP integration mechanics. Purpose This agent’s job is narrowly scoped: validate the CSV payload from SAP against externally stored business rules and produce outputs that the workflow can use deterministically. In other words, it turns “validation as reasoning” into workflow artifacts (summary + invalid IDs + invalid payload), instead of leaving validation as unstructured prose. In Azure Logic Apps terms, this is an agent loop: an iterative process where an LLM follows instructions and selects from available tools to complete a multi-step task. Logic Apps agent workflows explicitly support this “agent chooses tools to complete tasks” model (see Agent Workflows Concepts). Tools In Logic Apps agent workflows, a tool is a named sequence that contains one or more actions the agent can invoke to accomplish part of its task (see Agent Workflows Concepts). In the screenshot, the agent is configured with three tools, explicitly labeled Get validation rules, Get CSV payload, Summarize CSV payload review. These tool names match the business process in the “Instructions for agent” box (steps 1–3). The next sections of the post go deeper into what each tool does internally; at this level, the important point is simply that the agent is constrained to a small, explicit toolset. Agent execution The screenshot shows the agent configured with: AI model: gpt-5-3 (gpt-5) A connection line: “Connected to … (Azure OpenAI)” Instructions for agent that define the agent’s role and a 3-step business process: Get validation rules (via the ValidationRules reference) Get CSV payload Summarize the CSV payload review, using the validation document This pattern is intentional: The instructions provide the agent’s “operating procedure” in plain language. The tools give the agent: controlled ways to fetch the rule document, access the CSV input, and return structured results. Because the workflow consumes the agent’s outputs downstream, the instruction text is effectively part of your workflow contract (it must remain stable enough that later actions can trust the output shape). Note: If a reader wants to recreate this pattern, the fastest path is: Start with the official overview of agent workflows (Workflows with AI Agents and Models - Azure Logic Apps). Follow a hands-on walkthrough for building an agent workflow and connecting it to an Azure OpenAI deployment (Logic Apps Labs is a good step-by-step reference). [ azure.github.io ] Use the Azure OpenAI connector reference to understand authentication options and operations available in Logic Apps Standard (see Built-in OpenAI Connector) If you’re using Foundry for resource management, review how Foundry connections are created and used, especially when multiple resources/tools are involved (see How to connect to AI foundry). 2.1 Tool 1: Get validation rules The first tool in the validation agent loop is Get validation rules. Its job is to load the business validation rules that will be applied to the incoming CSV payload from SAP. I keep these rules outside the workflow (in a document) so they can be updated without redeploying the Logic App. In this example, the rules are stored in SharePoint, and the tool simply retrieves the document content at runtime. Get validation rules is implemented as a single action called Get validation document. In the designer, you can see it: uses a SharePoint Online connection (SharePoint icon and connector action) calls GetFileContentByPath (shown by the “File Path” input) reads the rule file from the configured Site Address uses the workflow parameter token ValidationRules for the File Path (so the exact rule file location is configurable per environment) The output of this tool is the raw rule document content, which the Data Validation Agent uses in the next steps to validate the CSV payload. The bottom half of the figure shows an excerpt of the rules document. The format is simple and intentionally human-editable: each rule is expressed as FieldName : condition. For example, the visible rules include: PaymentMethod : value must exist PaymentMethod : value cannot be “Cash” OrderStatus : value must be different from “Cancelled” CouponCode : value must have at least 1 character OrderID : value must be unique in the CSV array A scope note: “Do not validate the Date field.” These rules are the “source of truth” for validation. The workflow does not hardcode them into expressions; instead, it retrieves them from SharePoint and passes them into the agent loop so the validation logic remains configurable and auditable (you can always point to the exact rule document used for a given run). A small but intentional rule in the document is “Do not validate the Date field.” That line is there for a practical reason: in an early version of the source workflow, the date column was being corrupted during CSV generation. The validation agent still tried to validate dates (even though date validation wasn’t part of the original intent), and the result was predictable: every row failed validation, leaving nothing to analyze. The upstream issue is fixed now, but I kept this rule in the demo to illustrate an important point: validation is only useful when it’s aligned with the data contract you can actually guarantee at that point in the pipeline. Note: The rules shown here assume the CSV includes a header row (field names in the first line) so the agent can interpret each column by name. If you want the agent to be schema‑agnostic, you can extend the rules with an explicit column mapping, for example: Column 1: Order ID Column 2: Date Column 3: Customer ID … This makes the contract explicit even when headers are missing or unreliable. With the rules loaded, the next tool provides the second input the agent needs: the CSV payload that will be validated against this document. 2.2 Tool 2: Get CSV payload The second tool in the validation agent loop is Get CSV payload. Its purpose is to make the dataset-to-validate explicit: it defines exactly what the agent should treat as “the CSV payload,” rather than relying on implicit workflow context. In this workflow, the CSV is already constructed earlier (as Create_CSV_payload ), and this tool acts as the narrow bridge between that prepared string and the agent’s validation step. Figure: Tool #2 (“Get CSV payload”) defines a single agent parameter and binds it to the workflow’s generated CSV. The figure shows two important pieces: - The tool parameter contract (“Agent Parameters”) On the right, the tool defines an agent parameter named CSV Payload with type String, and the description (highlighted in yellow) makes the intent explicit: “The CSV payload received from SAP and that we validate based on the validation rules.” This parameter is the tool’s interface: it documents what the agent is supposed to provide/consume when using this tool, and it anchors the rest of the validation process to a single, well-defined input. Tools in Logic Apps agent workflows exist specifically to constrain and structure what an agent can do and what data it operates on (see Agent Workflows Concepts). - Why there is an explicit Compose action (“CSV payload”) In the lower-right “Code view,” the tool’s internal action is shown as a standard Compose: { "type": "Compose", "inputs": "@outputs('Create_CSV_payload')" } This is intentional. Even though the CSV already exists in the workflow, the tool still needs a concrete action that produces the value it returns to the agent. The Compose step: pins the tool output to a single source of truth ( Create_CSV_payload ), and creates a stable boundary: “this is the exact CSV string the agent validates,” independent of other workflow state. Put simply: the Compose action isn’t there because Logic Apps can’t access the CSV—it’s there to make the agent/tool interface explicit, repeatable, and easy to troubleshoot. What “tool parameters” are (in practical terms) In Logic Apps agent workflows, a tool is a named sequence of one or more actions that the agent can invoke while executing its instructions. A tool parameter is the tool’s input/output contract exposed to the agent. In this screenshot, that contract is defined under Agent Parameters, where you specify: Name: CSV Payload Type: String Description: “The CSV payload received from SAP…” This matters because it clarifies (for both the model and the human reader) what the tool represents and what data it is responsible for supplying. With Tool #1 providing the rules document and Tool #2 providing the CSV dataset, Tool #3 is where the agent produces workflow-ready outputs (summary + invalid IDs + filtered payload) that the downstream steps can act on. 2.3 Tool 3: Summarize CSV payload review The third tool, Summarize CSV payload review, is where the agent stops being “an evaluator” and becomes a producer of workflow-ready outputs. It does most of the heavy lifting so let's go into the details. Instead of returning one blob of prose, the tool defines three explicit agent parameters—each with a specific format and purpose—so the workflow can reliably consume the results in downstream actions. In Logic Apps agent workflows, tools are explicitly defined tasks the agent can invoke, and each tool can be structured around actions and schemas that keep the loop predictable (see Agent Workflows Concepts). Figure: Tool #3 (“Summarize CSV payload review”) defines three structured agent outputs Description is not just documentation—it’s the contract the model is expected to satisfy, and it strongly shapes what the agent considers “relevant” when generating outputs. The parameters are: Validation summary (String) Goal: a human-readable summary that can be dropped straight into email. In the screenshot, the description is very explicit about shape and content: “expected format is an HTML table” “create a list of all orderids that have failed” “create a CSV document… only for the orderid values that failed… each row on a separate line” “include title row only in the email body” This parameter is designed for presentation: it’s the thing you want humans to read first. InvalidOrderIds (String, CSV format) Goal: a machine-friendly list of identifiers the workflow can use deterministically. The key part of the description (highlighted in the image) is: “The format is CSV.” That single sentence is doing a lot of work: it tells the model to emit a comma-separated list, which you then convert into an array in the workflow using split(...). Invalid CSV payload (String, one row per line) Goal: the failed rows extracted from the original dataset, in a form that downstream steps can reuse. The description constrains the output tightly: “original CSV rows… for the orderid values that failed validation” “each row must be on a separate line” “keep the title row only for the email body and remove it otherwise” This parameter is designed for automation: it becomes input to remediation steps (like transforming rows to XML and creating IDocs), not just a report. What “agent parameters” do here (and why they matter) A useful way to think about agent parameters is: they are the “typed return values” of a tool. Tools in agent workflows exist to structure work into bounded tasks the agent can perform, and a schema/parameter contract makes the results consumable by the rest of the workflow (see Agent Workflows Concepts). In this tool, the parameters serve two purposes at once: They guide the agent toward salient outputs. The descriptions explicitly name what matters: “failed orderids,” “HTML table,” “CSV format,” “one row per line,” “header row rules.” That phrasing makes it much harder for the model to “wander” into irrelevant commentary. They align with how the workflow will parse and use the results. By stating “ InvalidOrderIds is CSV,” you make it trivially parseable (split), and by stating “Invalid CSV payload is one row per line,” you make it easy to feed into later transformations. Why the wording works (and what wording tends to work best) What’s interesting about the parameter descriptions is that they combine three kinds of constraints: Output format constraints (make parsing deterministic) “expected format is an HTML table” “The format is CSV.” “each row must be on a separate line” These format cues help the agent decide what to emit and help you avoid brittle parsing later. Output selection constraints (force relevance) “only for the orderid values that failed validation” “Create a list of all orderids that have failed” This tells the agent what to keep and what to ignore. Output operational constraints (tie outputs to downstream actions) “Include title row only in the email body” “remove it otherwise” This explicitly anticipates downstream usage (email vs remediation), which is exactly the kind of detail models often miss unless you state it. Rule of thumb: wording works best when it describes what to produce, in what format, with what filtering rules, and why the workflow needs it. How these parameters tie directly to the downstream actions The next picture makes the design intent very clear: each parameter is immediately “bound” to a normal workflow value via Compose actions and then used by later steps. This is the pattern we want: agent output → Compose → (optional) normalization → reused by deterministic workflow actions. It’s the opposite of “read the model output and hope.” This is the reusable pattern: Decide the minimal set of outputs the workflow needs. Specify formats that are easy to parse. Write parameter descriptions that encode both selection and formatting constraints. Immediately bind outputs to workflow variables via Compose/ SetVariable actions. The main takeaway from this tool is that the agent is being forced into a structured contract: three outputs with explicit formats and clear intent. That contract is what makes the rest of the workflow deterministic—Compose actions can safely read @agentParameters(...), the workflow can safely split(...) the invalid IDs, and downstream actions can treat the “invalid payload” as real data rather than narrative. I'll show later how this same “parameter-first” design scales to other agent tools. 2.4 Turning agent outputs into a verification email Once the agent has produced structured outputs (Validation summary, InvalidOrderIds , and Invalid CSV payload), the next goal is to make those outputs operational: humans need a quick summary of what failed, and the workflow needs machine‑friendly values it can reuse downstream. The design here is intentionally straightforward: the workflow converts each agent parameter into a first‑class workflow output (via Compose actions and one variable assignment), then binds those values directly into the Office 365 email body. The result is an email that is both readable and actionable—without anyone needing to open run history. The figure below shows how the outputs of Summarize CSV payload review are mapped into the verification email. On the left, the tool produces three values via subsequent actions (Summary, Invalid order ids, and Invalid CSV payload), and the workflow also normalizes the invalid IDs into an array (Save invalid order ids). On the right, the Send verification summary action composes the email body using those same values as dynamic content tokens. Figure: Mapping agent outputs to the verification email The important point is that the email is not constructed by “re-prompting” or “re-summarizing.” It is assembled from already-structured outputs. This mapping is intentionally direct: each piece of the email corresponds to one explicit output from the agent tool. The workflow doesn’t interpret or transform the summary beyond basic formatting—its job is to preserve the agent’s structured outputs and present them consistently. The only normalization step happens for InvalidOrderIds , where the workflow also converts the CSV string into an array ( ArrayOfInvalidOrderIDs ) for later filtering and analysis steps. The next figure shows a sample verification email produced by this pipeline. It illustrates the three-part structure: an HTML validation summary table, the raw invalid order ID list, and the extracted invalid CSV rows: Figure: Sample verification email — validation summary table + invalid order IDs + invalid CSV rows. The extracted artifacts InvalidOrderIds and Invalid CSV payload are used in the downstream actions that persist failed rows as IDocs for later processing, which were presented in Part 1. I will get back to this later to talk about reusing the validation agent. Next however I will go over the data analysis part of the AI integration. 3. Analysis Phase: from validated dataset to HTML output After the validation agent loop finishes, the workflow enters a second AI phase: analysis. The validation phase is deliberately about correctness (what to exclude and why). The analysis phase is about insight, and it runs on the remaining dataset after invalid rows are filtered out. At a high level, this phase has three steps: Call Azure OpenAI to analyze the CSV dataset while explicitly excluding invalid OrderIDs . Extract the model’s text output from the OpenAI response object. Convert the model’s markdown output into HTML so it renders cleanly in email (and in the SAP response envelope). 3.1 OpenAI component: the “Analyze data” call The figure below shows the Analyze data action that drives the analysis phase. This action is executed after the Data Validation Agent completes, and it uses three messages: a system instruction that defines the task, the CSV dataset as input, and a second user message that enumerates the OrderIDs to exclude (the invalid IDs produced by validation). Figure: Azure OpenAI analysis call. The analysis call is structured as: system: define the task and constraints user: provide the dataset user: provide exclusions derived from validation system: Analyze dataset; provide trends/predictions; exclude specified orderids. user: <csv payload=""> user: Excluded orderids: <comma-separated ids="" invalid=""></comma-separated></csv> Two design choices are doing most of the work here: The model is given the dataset and the exclusions separately. This avoids ambiguity: the dataset is one message, and the “do not include these OrderIDs ” constraint is another. The exclusion list is derived from validation output, not re-discovered during analysis. The analysis step doesn’t re-validate; it consumes the validation phase’s results and focuses purely on trends/predictions. 3.2 Processing the response The next figure shows how the workflow turns the Azure OpenAI response into a single string that can be reused for email and for the SAP response. The workflow does three things in sequence: it parses the response JSON, extracts the model’s text content, and then passes that text into an HTML formatter. Figure: Processing the OpenAI response. This is the only part of the OpenAI response you need to understand for this workflow: Analyze_data response └─ choices[] (array) └─ [0] (object) └─ message (object) └─ content (string) <-- analysis text Everything else in the OpenAI response (filters, indexes, metadata) is useful for auditing but not required to build the final user-facing output. 3.3 Crafting the output to HTML The model’s output is plain text and often includes lightweight markdown structures (headings, lists, separators). To make the analysis readable in email (and safe to embed in the SAP response envelope), the workflow converts the markdown into HTML. The script was generated with copilot. Source code snippet may be found in Part 1. The next figure shows what the formatted analysis looks like when rendered. Not the explicit reference to the excluded OrderIDs and summary of the remaining dataset before listing trend observations. Figure: Example analysis output after formatting. 4. Closing the loop: persisting invalid rows as IDocs In Part 1, I introduced an optional remediation branch: when validation finds bad rows, the workflow can persist them into SAP as custom IDocs for later handling. In Part 2, after unpacking the agent loop, I want to reconnect those pieces and show the “end of the story”: the destination workflow creates IDocs for invalid data, and a second destination workflow receives those IDocs and produces a consolidated audit trail in Blob Storage. This final section is intentionally pragmatic. It shows: where the IDoc creation call happens, how the created IDocs arrive downstream, and how to safely handle many concurrent workflow instances writing to the same storage artifact (one instance per IDoc). 4.1 From “verification summary” to “Create all IDocs” The figure below shows the tail end of the verification summary flow. Once the agent produces the structured validation outputs, the workflow first emails the human-readable summary, then converts the invalid CSV rows into an SAP-friendly XML shape, and finally calls the RFC that creates IDocs from those rows. Figure: End of the validation/remediation branch. This is deliberately a “handoff point.” After this step, the invalid rows are no longer just text in an email—they become durable SAP artifacts (IDocs) that can be routed, retried, and processed independently of the original workflow run. 4.2 Z_CREATE_ONLINEORDER_IDOC and the downstream receiver The next figure is the same overview from Part 1. I’m reusing it here because it captures the full loop: the workflow calls Z_CREATE_ONLINEORDER_IDOC , SAP converts the invalid rows into custom IDocs, and Destination workflow #2 receives those IDocs asynchronously (one workflow run per IDoc). Figure 2: Invalid rows persisted as custom IDocs. This pattern is intentionally modular: Destination workflow #1 decides which rows are invalid and optionally persists them. SAP encapsulates the IDoc creation mechanics behind a stable RFC ( Z_CREATE_ONLINEORDER_IDOC ). Destination workflow #2 processes each incoming IDoc independently, which matches how IDoc-driven integrations typically behave in production. 4.3 Two phases in Destination workflow #2: AI agent + Blob Storage logging In the receiver workflow, there are two distinct phases: AI agent phase (per-IDoc): reconstruct a CSV view from the incoming IDoc payload and (optionally) run the same validation logic. Blob storage phase (shared output): append a normalized “verification line” into a shared blob in a concurrency-safe way. It’s worth calling out: in this demo, the IDocs being received were created from already-validated outputs upstream, so you could argue the second validation is redundant. I keep it anyway for two reasons: it demonstrates that the agent tooling is reusable with minimal changes, and in a general integration, Destination workflow #2 may receive IDocs from multiple sources, not only from this pipeline—so “validate on receipt” can still be valuable. 4.3.1 AI agent phase The figure below shows the validation agent used in Destination workflow #2. The key difference from the earlier agent loop is the output format: instead of producing an HTML summary + invalid lists, this agent writes a single “audit line” that includes the IDoc correlation key ( DOCNUM ) along with the order ID and the failed rules. Figure: Destination workflow #2 agent configuration. The reusable part here is the tooling structure: rules still come from the same validation document, the dataset is still supplied as CSV, and the summarization tool outputs a structured value the workflow can consume deterministically. The only meaningful change is “what shape do I want the output to take,” which is exactly what the agent parameter descriptions control. The next figure zooms in on the summarization tool parameter in Destination workflow #2. Instead of three outputs, this tool uses a single parameter ( VerificationInfo ) whose description forces a consistent line format anchored on DOCNUM . Figure 4: VerificationInfo parameter. This is the same design principle as Tool #3 in the first destination workflow: describe the output as a contract, not as a vague request. The parameter description tells the agent exactly what must be present ( DOCNUM + OrderId + failed rules) and therefore makes it straightforward to append the output to a shared log without additional parsing. Interesting snippets Extracting DOCNUM from the IDoc control record and carry it through the run: xpath(xml(triggerBody()?['content']), 'string(/*[local-name()="Receive"] /*[local-name()="idocData"] /*[local-name()="EDI_DC40"] /*[local-name()="DOCNUM"])') 4.3.2 Blob Storage phase Destination workflow #2 runs one instance per inbound IDoc. That means multiple runs can execute at the same time, all trying to write to the same “ ValidationErrorsYYYYMMDD.txt ” artifact. The figure below shows the resulting appended output: one line per IDoc, each line beginning with DOCNUM , which becomes the stable correlation key. Destination workflow #2 runs one instance per inbound IDoc, so multiple instances can attempt to write to the same daily “validation errors” append blob at the same time. The figure below shows the concurrency control pattern I used to make those writes safe: a short lease acquisition loop that retries until it owns the blob lease, then appends the verification line(s), and finally releases the lease. Figure: Concurrency-safe append pattern. Reading the diagram top‑to‑bottom, the workflow uses a simple lease → append → release pattern to make concurrent writes safe. Each instance waits briefly (Delay), attempts to acquire a blob lease (Acquire validation errors blob lease), and loops until it succeeds (Set status code → Until lease is acquired). Once a lease is obtained, the workflow stores the lease ID (Save lease id), appends its verification output under that lease (Append verification results), and then releases the lease (Release the lease) so the next workflow instance can write. Implementation note: the complete configuration for this concurrency pattern (including the HTTP actions, headers, retries, and loop conditions) is included in the attached artifacts, in the workflow JSON for Destination workflow #2. 5. Concluding remarks Part 2 zoomed in on the AI boundary inside the destination workflows and made it concrete: what the agent sees, what it is allowed to do, what it must return, and how those outputs drive deterministic workflow actions. The practical outcomes of Part 2 are: A tool-driven validation agent that produces workflow artifacts, not prose. The validation loop is constrained by tools and parameter schemas so its outputs are immediately consumable: an email-friendly validation summary, a machine-friendly InvalidOrderIds list, and an invalid-row payload that can be remediated. A clean separation between validation and analysis. Validation decides what not to trust (invalid IDs / rows) and analysis focuses on what is interesting in the remaining dataset. The analysis prompt makes the exclusion rule explicit by passing the dataset and excluded IDs as separate messages. A repeatable response-processing pipeline. You extract the model’s text from a stable response path ( choices[0].message.content ), then shape it into HTML once (markdown → HTML) so the same formatted output can be reused for email and the SAP response envelope. A “reuse with minimal changes” pattern across workflows. Destination workflow #2 shows the same agent principles applied to IDoc reception, but with a different output contract optimized for logging: DOCNUM + OrderId + FailedRules . This demonstrates that the real reusable asset is the tool + parameter contract design. Putting It All Together We have a full integration story where SAP, Logic Apps, AI, and IDocs are connected with explicit contracts and predictable behavior: Part 1 established the deterministic integration foundation. SAP ↔ Logic Apps connectivity (gateway/program wiring) RFC payload/response contracts ( IT_CSV , response envelope, error semantics) predictable exception propagation back into SAP an optional remediation branch that persists invalid rows as IDocs via a custom RFC ( Z_CREATE_ONLINEORDER_IDOC ) and the end-to-end response handling pattern in the caller workflow. Part 2 layered AI on top without destabilizing the contracts. Agent loop + tools for rule retrieval and validation output schemas that convert “reasoning” into workflow artifacts a separate analysis step that consumes validated data and produces formatted results and an asynchronous IDoc receiver that logs outcomes safely under concurrency. The reason it works as a two-part series is that the two layers evolve at different speeds: The integration layer (Part 1) should change slowly. It defines interoperability: payload shapes, RFC names, error contracts, and IDoc interfaces. The AI layer (Part 2) is expected to iterate. Prompts, rule documents, output formatting, and agent tool design will evolve as you tune behavior and edge cases. References Logic Apps Agentic Workflows with SAP - Part 1: Infrastructure 🤖 Agent Loop Demos 🤖 | Microsoft Community Hub Agent Workflows Concepts Workflows with AI Agents and Models - Azure Logic Apps Built-in OpenAI Connector How to connect to AI foundry Create Autonomous AI Agent Workflows - Azure Logic Apps Handling Errors in SAP BAPI Transactions Access SAP from workflows Create common SAP workflows Generate Schemas for SAP Artifacts via Workflows Exception Handling | ABAP Keyword Documentation Handling and Propagating Exceptions - ABAP Keyword Documentation SAP .NET Connector 3.1 Overview SAP .NET Connector 3.1 Programming Guide Connect to Azure AI services from Workflows All supporting content for this post may be found in the companion GitHub repository.Logic Apps Agentic Workflows with SAP - Part 1: Infrastructure
When you integrate Azure Logic Apps with SAP, the “hello world” part is usually easy. The part that bites you later is data quality. In SAP-heavy flows, validation isn’t a nice-to-have — it’s what makes the downstream results meaningful. If invalid data slips through, it can get expensive fast: you may create incorrect business documents, trigger follow-up processes, and end up in a cleanup path that’s harder (and more manual) than building validation upfront. And in “all-or-nothing” transactional patterns, things get even more interesting: one bad record can force a rollback strategy, compensating actions, or a whole replay/reconciliation story you didn’t want to own. See for instance Handling Errors in SAP BAPI Transactions | Microsoft Community Hub to get an idea of the complexity in a BizTalk context. That’s the motivation for this post: a practical starter pattern that you can adapt to many data shapes and domains for validating data in a Logic Apps + SAP integration. Note: For the full set of assets used here, see the companion GitHub repository (workflows, schemas, SAP ABAP code, and sample files). 1. Introduction Scenario overview The scenario is intentionally simple, but it mirrors what shows up in real systems: A Logic App workflow sends CSV documents to an SAP endpoint. SAP forwards the payload to a second Logic App workflow that performs: rule-based validation (based on pre-defined rules) analysis/enrichment (market trends, predictions, recommendations) The workflow either: returns validated results (or validation errors) to the initiating workflow, or persists outputs for later use For illustration, I’m using fictitious retail data. The content is made up, but the mechanics are generic: the same approach works for orders, inventory, pricing, master data feeds, or any “file in → decision out” integration. You’ll see sample inputs and outputs below to keep the transformations concrete. What this post covers This walkthrough focuses on the integration building blocks that tend to matter in production: Calling SAP RFCs from Logic App workflows, and invoking Logic App workflows from SAP function modules Using the Logic Apps SAP built-in trigger Receiving and processing IDocs Returning responses and exceptions back to SAP in a structured, actionable way Data manipulation patterns in Logic Apps, including: parsing and formatting inline scripts XPath (where it fits, and where it becomes painful). Overall Implementation A high-level view of the implementation is shown below. The source workflow handles end-to-end ingestion—file intake, transformation, SAP integration, error handling, and notifications—using Azure Logic Apps. The destination workflows focus on validation and downstream processing, including AI-assisted analysis and reporting, with robust exception handling across multiple technologies. I’ll cover the AI portion in a follow-up post. Note on AI-assisted development Most of the workflow “glue” in this post—XPath, JavaScript snippets, and Logic Apps expressions—was built with help from Copilot and the AI assistant in the designer (see Get AI-assisted help for Standard workflows - Azure Logic Apps | Microsoft Learn). In my experience, this is exactly where AI assistance pays off: generating correct scaffolding quickly, then iterating based on runtime behavior. I’ve also included SAP ABAP snippets for the SAP-side counterpart. You don’t need advanced ABAP knowledge to follow along; the snippets are deliberately narrow and integration-focused. I include them because it’s hard to design robust integrations if you only understand one side of the contract. When you understand how SAP expects to receive data, how it signals errors, and where transactional boundaries actually are, you end up with cleaner workflows and fewer surprises. 2. Source Workflow This workflow is a small, end‑to‑end “sender” pipeline: it reads a CSV file from Azure Blob Storage, converts the rows into the SAP table‑of‑lines XML shape expected by an RFC, calls Z_GET_ORDERS_ANALYSIS via the SAP connector, then extracts analysis or error details from the RFC response and emails a single consolidated result. The name of the data file is a parameter of the logic app. At a high level: Input: an HTTP request (used to kick off the run) + a blob name. Processing: CSV → array of rows → XML (…) → RFC call Output: one email containing either: the analysis (success path), or a composed error summary (failure path). The diagram below summarizes the sender pipeline: HTTP trigger → Blob CSV (header included) → rows → SAP RFC → parse response → email. Two design choices are doing most of the work here. First, the workflow keeps the CSV transport contract stable by sending the file as a verbatim list of lines—including the header—wrapped into … elements under IT_CSV . Second, it treats the RFC response as the source of truth: EXCEPTIONMSG and RETURN/MESSAGE drive a single Has errors gate, which determines whether the email contains the analysis or a consolidated failure summary. Step-by-step description Phase 0 — Trigger Trigger — When_an_HTTP_request_is_received The workflow is invoked via an HTTP request trigger (stateful workflow). Phase 1 — Load and split the CSV Read file — Read_CSV_orders_from_blob Reads the CSV from container onlinestoreorders using the blob name from @parameters('DataFileName'). The name of the data file is a parameter of the logic app. Split into rows — Extract_rows Splits the blob content on \r\n, producing an array of CSV lines. Design note: Keeping the header row is useful when downstream validation or analysis wants column names, and it avoids implicit assumptions in the sender workflow. Phase 2 — Shape the RFC payload Convert CSV rows to SAP XML — Transform_CSV_to_XML Uses JavaScript to wrap each CSV row (including the header line) into the SAP line structure and XML‑escape special characters. The output is an XML fragment representing a table of ZTY_CSV_LINE rows. Phase 3 — Call SAP and extract response fields Call the RFC — [RFC]_ Z_GET_ORDERS_ANALYSIS Invokes Z_GET_ORDERS_ANALYSIS with an XML body containing … built from the transformed rows. Note that a <DEST>...</DEST> element can also be provided to override the default value in the function module definition. Extract error/status — Save_EXCEPTION_message and Save_RETURN_message Uses XPath to pull: EXCEPTIONMSG from the RFC response, and the structured RETURN / MESSAGE field. Phase 4 — Decide success vs failure and notify Initialize output buffer — Initialize_email_body Creates the EmailBody variable used by both success and failure cases. Gate — Has_errors Determines whether to treat the run as failed based on: EXCEPTIONMSG being different from "ok", or RETURN / MESSAGE being non‑empty. Send result — Send_an_email_(V2) Emails either: the extracted ANALYSIS (success), or a concatenated error summary including RETURN / MESSAGE plus message details ( MESSAGE_V1 … MESSAGE_V4 ) and EXCEPTIONMSG . Note: Because the header row is included in IT_CSV , the SAP-side parsing/validation treats the first line as column titles (or simply ignores it). The sender workflow stays “schema-agnostic” by design. Useful snippets Snippet 1 — Split the CSV into rows split(string(body('Read_CSV_orders_from_blob')?['content']), '\r\n') Tip: If your CSV has a header row you don’t want to send to SAP, switch back to: @skip(split(string(body('Read_CSV_orders_from_blob')?['content']), '\r\n'), 1) Snippet 2 — JavaScript transform: “rows → SAP table‑of‑lines XML” const lines = workflowContext.actions.Extract_rows.outputs; function xmlEscape(value) { return String(value) .replace(/&/g, "&") .replace(//g, ">") .replace(/"/g, """) .replace(/'/g, "'"); } // NOTE: we don't want to keep empty lines (which can be produced by reading the blobs) // the reason being that if the recipient uses a schema to validate the xml, // it may reject it if it does not allow empty nodes. const xml = lines .filter(line => line && line.trim() !== '') // keep only non-empty lines .map(line => `<zty_csv_line><line>${xmlEscape(line)}</line></zty_csv_line>`) .join(''); return { xml }; Snippet 3 — XPath extraction of response fields (namespace-robust) EXCEPTIONMSG: @xpath(body('[RFC]_Call_Z_GET_ORDERS_ANALYSIS')?['content'], 'string( /*[local-name()="Z_GET_ORDERS_ANALYSISResponse"] /*[local-name()="EXCEPTIONMSG"])') RETURN/MESSAGE: @xpath(body('[RFC]_Call_Z_GET_ORDERS_ANALYSIS')?['content'], 'string( /*[local-name()="Z_GET_ORDERS_ANALYSISResponse"] /*[local-name()="RETURN"] /*[local-name()="MESSAGE"])') Snippet 4 — Failure email body composition concat( 'Error message: ', outputs('Save_RETURN_message'), ', details: ', xpath(body('[RFC]_Call_Z_GET_ORDERS_ANALYSIS')?['content'], 'string(//*[local-name()=\"MESSAGE_V1\"])'), xpath(body('[RFC]_Call_Z_GET_ORDERS_ANALYSIS')?['content'], 'string(//*[local-name()=\"MESSAGE_V2\"])'), xpath(body('[RFC]_Call_Z_GET_ORDERS_ANALYSIS')?['content'], 'string(//*[local-name()=\"MESSAGE_V3\"])'), xpath(body('[RFC]_Call_Z_GET_ORDERS_ANALYSIS')?['content'], 'string(//*[local-name()=\"MESSAGE_V4\"])'), '; ', 'Exception message: ', outputs('Save_EXCEPTION_message'), '.') 3. SAP Support To make the SAP/Logic Apps boundary simple, I model the incoming CSV as a table of “raw lines” on the SAP side. The function module Z_GET_ORDERS_ANALYSIS exposes a single table parameter, IT_CSV , typed using a custom line structure. Figure: IT_CSV is a table of CSV lines ( ZTY_CSV_LINE ), with a single LINE field ( CHAR2048 ). IT_CSV uses the custom structure ZTY_CSV_LINE , which contains a single component LINE ( CHAR2048 ). This keeps the SAP interface stable: the workflow can send CSV lines without SAP having to know the schema up front, and the parsing/validation logic can evolve independently. The diagram below shows the plumbing that connects SAP to Azure Logic Apps in two common patterns: SAP sending IDocs to a workflow and SAP calling a remote-enabled endpoint via an RFC destination. I’m showing all three pieces together—the ABAP call site, the SM59 RFC destination, and the Logic Apps SAP built-in trigger—because most “it doesn’t work” problems come down to a small set of mismatched configuration values rather than workflow logic. The key takeaway is that both patterns hinge on the same contract: Program ID plus the SAP Gateway host/service. In SAP, those live in SM59 (TCP/IP destination, registered server program). In Logic Apps, the SAP built-in trigger listens using the same Program ID and gateway settings, while the trigger configuration (for example, IDoc format and degree of parallelism) controls how messages are interpreted and processed. Once these values line up, the rest of the implementation becomes “normal workflow engineering”: validation, predictable error propagation, and response shaping. Before diving into workflow internals, I make the SAP-side contract explicit. The function module interface below shows the integration boundary: CSV lines come in as IT_CSV , results come back as ANALYSIS , and status/error information is surfaced both as a human-readable EXCEPTIONMSG and as a structured RETURN ( BAPIRET2 ). I also use a dedicated exception ( SENDEXCEPTIONTOSAPSERVER ) to signal workflow-raised failures cleanly. Contract (what goes over RFC): Input: IT_CSV (CSV lines) Outputs: ANALYSIS (analysis payload), EXCEPTIONMSG (human-readable status) Return structure: RETURN ( BAPIRET2 ) for structured SAP-style success/error Custom exception: SENDEXCEPTIONTOSAPSERVER for workflow-raised failures Here is the ABAP wrapper that calls the remote implementation and normalizes the result. FUNCTION z_get_orders_analysis. *"---------------------------------------------------------------------- *" This module acts as a caller wrapper. *" Important: the remote execution is determined by DESTINATION. *" Even though the function name is the same, this is not recursion: *" the call runs in the remote RFC server registered under DESTINATION "DEST". *"---------------------------------------------------------------------- *" Contract: *" TABLES it_csv "CSV lines *" IMPORTING analysis "Result payload *" EXPORTING exceptionmsg "Human-readable status / error *" CHANGING return "BAPIRET2 return structure *" EXCEPTIONS sendexceptiontosapserver *"---------------------------------------------------------------------- CALL FUNCTION 'Z_GET_ORDERS_ANALYSIS' DESTINATION dest IMPORTING analysis = analysis TABLES it_csv = it_csv CHANGING return = return EXCEPTIONS sendexceptiontosapserver = 1 system_failure = 2 MESSAGE exceptionmsg communication_failure = 3 MESSAGE exceptionmsg OTHERS = 4. CASE sy-subrc. WHEN 0. exceptionmsg = 'ok'. "Optional: normalize success into RETURN for callers that ignore EXCEPTIONMSG IF return-type IS INITIAL. return-type = 'S'. return-message = 'OK'. ENDIF. WHEN 1. exceptionmsg = |Exception from workflow: SENDEXCEPTIONTOSAPSERVER { sy-msgv1 }{ sy-msgv2 }{ sy-msgv3 }{ sy-msgv4 }|. return-type = 'E'. return-message = exceptionmsg. WHEN 2 OR 3. "system_failure / communication_failure usually already populate exceptionmsg IF exceptionmsg IS INITIAL. exceptionmsg = |RFC system/communication failure.|. ENDIF. return-type = 'E'. return-message = exceptionmsg. WHEN OTHERS. exceptionmsg = |Error in workflow: { sy-msgv1 }{ sy-msgv2 }{ sy-msgv3 }{ sy-msgv4 }|. return-type = 'E'. return-message = exceptionmsg. ENDCASE. ENDFUNCTION. The wrapper is intentionally small: it forwards the payload to the remote implementation via the RFC destination and then normalizes the outcome into a predictable shape. The point isn’t fancy ABAP — it’s reliability. With a stable contract ( IT_CSV, ANALYSIS, RETURN, EXCEPTIONMSG ) the Logic Apps side can evolve independently while SAP callers still get consistent success/error semantics. Important: in 'CALL FUNCTION 'Z_GET_ORDERS_ANALYSIS' DESTINATION dest' the name of the called function should be the same as the name of the ABAP wrapper function module, the reason being that the SAP built-in trigger in the logic app uses the function module signature as the contract (i.e. metadata). Also note that in the provided companion artifacts, a default value was assigned to 'dest' in the function module definition. But it can also be specified in the request (DEST element). Z_GET_ORDERS_ANALYSIS . To sum up, the integration is intentionally shaped around three outputs: the raw input table ( IT_CSV ), a standardized SAP return structure ( RETURN / BAPIRET2 ), and a readable status string ( EXCEPTIONMSG ). The custom exception ( SENDEXCEPTIONTOSAPSERVER ) gives me a clean way to surface workflow failures back into SAP without burying them inside connector-specific error payloads. This is depicted in the figure below. 4. Destination Workflow The diagram below shows the destination workflow at a high level. I designed it as a staged pipeline: guard early, normalize input, validate, and then split the workload into two paths—operational handling of invalid records (notifications and optional IDoc remediation) and analysis of the validated dataset. Importantly, the SAP response is intentionally narrow: SAP receives only the final analysis (or a structured error), while validation details are delivered out-of-band via email. How to read this diagram Guardrail: Validate requested action ensures the workflow only handles the expected request. Normalize: Create CSV payload converts the inbound content into a consistent CSV representation. Validate: Data Validation Agent identifies invalid records (and produces a summary). Operational handling (invalid data): invalid rows are reported by email and may optionally be turned into IDocs (right-hand block). Analyze (valid data): Analyze data runs only on the validated dataset (invalid IDs excluded). Outputs: users receive Email analysis, while SAP receives only the analysis (or a structured error) via Respond to SAP server. Figure: Destination workflow with staged validation, optional IDoc remediation, and an SAP response . Reading the workflow top-to-bottom, the main design choice is separation of concerns. Validation is used to filter and operationalize bad records (notify humans, optionally create IDocs), while the SAP-facing response stays clean and predictable: SAP receives the final analysis for the validated dataset, or an error if the run can’t complete. This keeps the SAP contract stable even as validation rules and reporting details evolve. Step‑by‑step walkthrough Phase 0 — Entry and routing Trigger — When a message is received The workflow starts when an inbound SAP message is delivered to the Logic Apps SAP built‑in trigger. Guardrail — Validate requested action (2 cases) The workflow immediately checks whether the inbound request is the operation it expects (for example, the function/action name equals Z_GET_ORDERS_ANALYSIS ). If the action does not match: the workflow sends an exception back to SAP describing the unexpected action and terminates early (fail fast). If the action matches: processing continues. Phase 1 — Normalize input into a workflow‑friendly payload Prepare input — Create CSV payload The workflow extracts CSV lines from the inbound (XML) SAP payload and normalizes them into a consistent CSV text payload that downstream steps can process reliably. Initialize validation state — Initialize array of invalid order ids The workflow creates an empty array variable to capture order IDs that fail validation. This becomes the “validation output channel” used later for reporting, filtering, and optional remediation. Phase 2 — Validate the dataset (AI agent loop) Validate — Data Validation Agent (3 cases) This stage performs rule‑based validation using an agent pattern (backed by Azure OpenAI). Conceptually, it does three things (as shown in the diagram’s expanded block): Get validation rules: retrieves business rules from a SharePoint‑hosted validation document. The location of the validation rules file is a parameter of the logic app. Get CSV payload: loads the normalized CSV created earlier. Summarize CSV payload review: evaluates the CSV against the rules and produces structured validation outputs. Outputs produced by validation: A list of invalid order IDs The corresponding invalid CSV rows A human‑readable validation summary Note: The detailed AI prompt/agent mechanics are covered in Part 2. In Part 1, the focus is on the integration flow and how data moves. Phase 3 — Operational handling of invalid records (email + optional SAP remediation) After validation, the workflow treats invalid records as an operational concern: they are reported to humans and can optionally be routed into an SAP remediation path. This is shown in the right‑hand “Create IDocs” block. Notify — Send verification summary The workflow sends an email report (Office 365) to configured recipients containing: the validation summary the invalid order IDs the invalid CSV payload (or the subset of invalid rows) Transform — Transform CSV to XML The workflow converts the invalid CSV lines into an XML format that is suitable for SAP processing. Optional remediation — [RFC] Create all IDocs (conditional) If the workflow parameter (for example, CreateIDocs) is enabled, the workflow calls an SAP RFC (e.g., Z_CREATE_ONLINEORDER_IDOC ) to create IDocs from the transformed invalid data. Why this matters: Validation results are made visible (email) and optionally actionable (IDocs), without polluting the primary analysis response that SAP receives. Phase 4 — Analyze only the validated dataset (AI analysis) The workflow runs AI analysis on the validated dataset, explicitly excluding invalid order IDs discovered during the validation phase. The analysis prompt instructs the model to produce outputs such as trends, predictions, and recommendations. Note: The AI analysis prompt design and output shaping are covered in Part 2. Phase 5 — Post‑process the AI response and publish outputs Package results — Process analysis results (Scope) The workflow converts the AI response into a format suitable for email and for SAP consumption: Parse the OpenAI JSON response Extract the analysis content Convert markdown → HTML using custom JavaScript formatting Outputs Email analysis: sends the formatted analysis to recipients. Respond to SAP server: returns only the analysis (and errors) to SAP. Key design choice: SAP receives a clean, stable contract—analysis on success, structured error on failure. Validation details are handled out‑of‑band via email (and optionally via IDoc creation). Note: the analysis email sent by the destination workflow is there for testing purposes, to verify that the html content remains the same as it is sent back to the source workflow. Useful snippets Snippet 1 - Join each CSV line in the XML to make a CSV table: join( xpath( xml(triggerBody()?['content']), '/*[local-name()=\"Z_GET_ORDERS_ANALYSIS\"] /*[local-name()=\"IT_CSV\"] /*[local-name()=\"ZTY_CSV_LINE\"] /*[local-name()=\"LINE\"]/text()' ), '\r\n') Note: For the sake of simplicity, XPath is used here and throughout all places where XML is parsed. In the general case however, the Parse XML with schema action is the better and recommended way to strictly enforce the data contract between senders and receivers. More information about Parse XML with schema is provided in Appendix 1. Snippet 2 - Format markdown to html (simplified): const raw = workflowContext.actions.Extract_analysis.outputs; // Basic HTML escaping for safety (keeps <code> blocks clean) const escapeHtml = s => s.replace(/[&<>"]/g, c => ({'&':'&','<':'<','>':'>','"':'"'}[c])); // Normalize line endings let md = raw; // raw.replace(/\r\n/g, '\n').trim(); // Convert code blocks (``` ... ```) md = md.replace(/```([\s\S]*?)```/g, (m, p1) => `<pre><code>${escapeHtml(p1)}</code></pre>`); // Horizontal rules --- or *** md = md.replace(/(?:^|\n)---+(?:\n|$)/g, '<hr/>'); // Headings ###### to # for (let i = 6; i >= 1; i--) { const re = new RegExp(`(?:^|\\n)${'#'.repeat(i)}\\s+(.+?)\\s*(?=\\n|$)`, 'g'); md = md.replace(re, (m, p1) => `<h${i}>${p1.trim()}</h${i}>`); } // Bold and italic md = md.replace(/\*\*([^*]+)\*\*/g, '<strong>$1</strong>'); md = md.replace(/\*([^*]+)\*/g, '<em>$1</em>'); // Unordered lists (lines starting with -, *, +) md = md.replace(/(?:^|\n)([-*+]\s.+(?:\n[-*+]\s.+)*)/g, (m) => { const items = m.trim().split(/\n/).map(l => l.replace(/^[-*+]\s+/, '').trim()); return '\n<ul>' + items.map(i => `<li>${i}</li>`).join('\n') + '</ul>'; }); // Paragraphs: wrap remaining text blocks in <p>...</p> const blocks = md.split(/\n{2,}/).map(b => { if (/^<h\d>|^<ul>|^<pre>|^<hr\/>/.test(b.trim())) return b; return `<p>${b.replace(/\n/g, '<br/>')}</p>`; }); const html = blocks.join(''); return { html }; 5. Exception Handling To illustrate exception handling, we supposed that multiple workflows may listen to the same program id (by design or unexpectedly) and could therefore receive messages that were meant for others. So the first thing that happens is validate that the function name is as expected. It is shown below. In this section I show three practical ways to surface workflow failures back to SAP using the Logic Apps action “Send exception to SAP server”, and the corresponding ABAP patterns used to handle them. The core idea is the same in all three: Logic Apps raises an exception on the SAP side, SAP receives it as an RFC exception, and your ABAP wrapper converts that into something predictable (for example, a readable EXCEPTIONMSG , a populated RETURN , or both). The differences are in how much control you want over the exception identity and whether you want to leverage SAP message classes for consistent, localized messages. 5.1 Default exception This first example shows the default behavior of Send exception to SAP server. When the action runs without a custom exception name configuration, the connector raises a pre-defined exception that can be handled explicitly in ABAP. On the Logic Apps side, the action card “Send exception to SAP server” sends an Exception Error Message (for example, “Unexpected action in request: …”). On the ABAP side, the RFC call lists SENDEXCEPTIONTOSAPSERVER = 1 under EXCEPTIONS , and the code uses CASE sy-subrc to map that exception to a readable message. The key takeaway is that you get a reliable “out-of-the-box” exception path: ABAP can treat sy-subrc = 1 as the workflow‑raised failure and generate a consistent EXCEPTIONMSG . This is the simplest option and works well when you don’t need multiple exception names—just one clear “workflow failed” signal. 5.2 Message exception If you want more control than the default, you can configure the action to raise a named exception declared in your ABAP function module interface. This makes it easier to route different failure types without parsing free-form text. The picture shows Advanced parameters under the Logic Apps action, including “Exception Name” with helper text indicating it must match an exception declared in the ABAP function module definition. This option is useful when you want to distinguish workflow error categories (e.g., validation vs. routing vs. downstream failures) using exception identity, not just message text. The contract stays explicit: Logic Apps raises a named exception, and ABAP can branch on that name (or on sy-subrc mapping) with minimal ambiguity. 5.3 Message class exception The third approach uses SAP’s built-in message class mechanism so that the exception raised by the workflow can map cleanly into SAP’s message catalog ( T100 ). This is helpful when you want consistent formatting and localization aligned with standard SAP patterns. On the Logic Apps side, the action shows advanced fields including Message Class, Message Number, and an Is ABAP Message toggle, with helper text stating the message class can come from message maintenance ( SE91 ) or be custom. On the ABAP side, the code highlights an error-handling block that calls using sy-msgid , sy-msgno , and variables sy-msgv1 … sy-msgv4 , then stores the resulting text in EXCEPTIONMSG . This pattern is ideal when you want workflow exceptions to look and behave like “native” SAP messages. Instead of hard-coding strings, you rely on the message catalog and let ABAP produce a consistent final message via FORMAT_MESSAGE . The result is easier to standardize across teams and environments—especially if you already manage message classes as part of your SAP development process. Refer to Appendix 2 for further information on FORMAT_MESSAGE . 5.4 Choosing an exception strategy that SAP can act on Across these examples, the goal is consistent: treat workflow failures as first‑class outcomes in SAP, not as connector noise buried in run history. The Logic Apps action Send exception to SAP server gives you three increasingly structured ways to do that, and the “right” choice depends on how much semantics you want SAP to understand. Default exception (lowest ceremony): Use this when you just need a reliable “workflow failed” signal. The connector raises a pre-defined exception name (for example, SENDEXCEPTIONTOSAPSERVER ), and ABAP can handle it with a simple EXCEPTIONS … = 1 mapping and a sy-subrc check. This is the fastest way to make failures visible and deterministic. Named exception(s) (more routing control): Use this when you want SAP to distinguish failure types without parsing message text. By raising an exception name declared in the ABAP function module interface, you can branch cleanly in ABAP (or map to different return handling) and keep the contract explicit and maintainable. Message class + number (most SAP-native): Use this when you want errors to look and behave like standard SAP messages—consistent wording, centralized maintenance, and better alignment with SAP operational practices. In this mode, ABAP can render the final localized string using FORMAT_MESSAGE and return it as EXCEPTIONMSG (and optionally BAPIRET2 - MESSAGE ), which makes the failure both human-friendly and SAP-friendly. A practical rule of thumb: start with the default exception while you stabilize the integration, move to named exceptions when you need clearer routing semantics, and adopt message classes when you want SAP-native error governance (standardization, maintainability, and localization). Regardless of the option, the key is to end with a predictable SAP-side contract: a clear success path, and a failure path that produces a structured return and a readable message. 6. Response Handling This section shows how the destination workflow returns either a successful analysis response or a workflow exception back to SAP, and how the source (caller) workflow interprets the RFC response structure to produce a single, human‑readable outcome (an email body). The key idea is to keep the SAP-facing contract stable: SAP always returns a Z_GET_ORDERS_ANALYSISResponse envelope, and the caller workflow decides between success and error using just two fields: EXCEPTIONMSG and RETURN / MESSAGE . To summarize the steps: Destination workflow either: sends a normal response via Respond to SAP server, or raises an exception via Send exception to SAP server (with an error message). SAP server exposes those outcomes through the RFC wrapper: sy-subrc = 0 → success ( EXCEPTIONMSG = 'ok') sy-subrc = 1 → workflow exception ( SENDEXCEPTIONTOSAPSERVER ) sy-subrc = 2/3 → system/communication failures Source workflow calls the RFC, extracts: EXCEPTIONMSG RETURN / MESSAGE and uses an Has errors gate to choose between a success email body (analysis) or a failure email body (error summary). The figure below shows the full return path for results and failures. On the right, the destination workflow either responds normally (Respond to SAP server) or raises a workflow exception (Send exception to SAP server). SAP then maps that into the RFC outcome ( sy-subrc and message fields). On the left, the source workflow parses the RFC response structure and populates a single EmailBody variable using two cases: failure (error details) or success (analysis text). Figure: Response/exception flow Two things make this pattern easy to operationalize. First, the caller workflow does not need to understand every SAP field—only EXCEPTIONMSG and RETURN / MESSAGE are required to decide success vs failure. Second, the failure path intentionally aggregates details ( MESSAGE_V1 … MESSAGE_V4 plus the exception text) into a single readable string so errors don’t get trapped in run history. Callout: The caller workflow deliberately treats EXCEPTIONMSG != "ok" or RETURN / MESSAGE present as the single source of truth for failure, which keeps the decision logic stable even if the response schema grows. Detailed description Phase 1 — Destination workflow: choose “response” vs “exception” Respond to SAP server returns the normal response payload back to SAP. Send exception to SAP server raises a workflow failure with an Exception Error Message (the screenshot shows an example beginning with “Unexpected action in request:” and a token for Function Name). Outcome: SAP receives either a normal response or a raised exception for the RFC call. Phase 2 — SAP server: map workflow outcomes to RFC results The SAP-side wrapper code shown in the figure calls: CALL FUNCTION ' Z_GET_ORDERS_ANALYSIS ' DESTINATION DEST ... It declares exception mappings including: SENDEXCEPTIONTOSAPSERVER = 1 system_failure = 2 MESSAGE EXCEPTIONMSG communication_failure = 3 MESSAGE EXCEPTIONMSG OTHERS = 4 Then it uses CASE sy-subrc . to normalize outcomes (the figure shows WHEN 0. setting EXCEPTIONMSG = 'ok'., and WHEN 1. building a readable message for the workflow exception). Outcome: regardless of why it failed, SAP can provide a consistent set of fields back to the caller: a return structure and an exception/status message. Phase 3 — Source workflow: parse response and build one “email body” After the RFC action ([RFC] Call Z GET ORDERS ANALYSIS ) the source workflow performs: Save EXCEPTION message Extracts EXCEPTIONMSG from the response XML using XPath. Save RETURN message Extracts RETURN / MESSAGE from the response XML using XPath. Initialize email body Creates EmailBody once, then sets it in exactly one of two cases. Has errors (two cases) The condition treats the run as “error” if either: EXCEPTIONMSG is not equal to "ok", or RETURN / MESSAGE is not empty. Set email body (failure) / Set email body (success) Failure: builds a consolidated string containing RETURN / MESSAGE , message details ( MESSAGE_V1 ..V4), and EXCEPTIONMSG . Success: sets EmailBody to the ANALYSIS field extracted from the response. Outcome: the caller produces a single artifact (EmailBody) that is readable and actionable, without requiring anyone to inspect the raw RFC response. Note: the email recipient is set as a logic app parameter. 7. Destination Workflow #2: Persisting failed rows as custom IDocs In this section I zoom in on the optional “IDoc persistence” branch at the end of the destination workflow. After the workflow identifies invalid rows (via the Data Validation Agent) and emails a verification summary, it can optionally call a second SAP RFC to save the failed rows as IDocs for later processing. This is mainly included to showcase another common SAP integration scenario—creating/handling IDocs—and to highlight that you can combine “AI-driven validation” with traditional enterprise workflows. The deeper motivation for invoking this as part of the agent tooling is covered in Part 2; here, the goal is to show the connector pattern and the custom RFC used to create IDocs from CSV input. The figure below shows the destination workflow at two levels: a high-level overview at the top, and a zoomed view of the post-validation remediation steps at the bottom. The zoom starts from Data Validation Agent → Summarize CSV payload review and then expands the sequence that runs after Send verification summary: Transform CSV to XML followed by an SAP RFC call that creates IDocs from the failed data. The key point is that this branch is not the main “analysis response” path. It’s a practical remediation option: once invalid rows are identified and reported, the workflow can persist them into SAP using a dedicated RFC ( Z_CREATE_ONLINEORDER_IDOC ) and a simple IT_CSV payload. This keeps the end-to-end flow modular: analysis can remain focused on validated data, while failed records can be routed to SAP for follow-up processing on their own timeline. Callout: This branch exists to showcase an IDoc-oriented connector scenario. The “why this is invoked from the agent tooling” context is covered in Part 2; here the focus is the mechanics of calling Z_CREATE_ONLINEORDER_IDOC with IT_CSV and receiving ET_RETURN / ET_DOCNUMS . The screenshot shows an XML body with the RFC root element and an SAP namespace: <z_create_onlineorder_idoc xmlns="http://Microsoft.LobServices.Sap/2007/03/Rfc/"> <iv_direction>...</iv_direction> <iv_sndptr>...</iv_sndptr> <iv_sndprn>...</iv_sndprn> <iv_rcvptr>...</iv_rcvptr> <iv_rcvprn>...</iv_rcvprn> <it_csv> @{ ...Outputs... } </it_csv> <et_return></et_return> <et_docnums></et_docnums> </z_create_onlineorder_idoc> What to notice: the workflow passes invalid CSV rows in IT_CSV , and SAP returns a status table ( ET_RETURN ) and created document numbers ( ET_DOCNUMS ) for traceability. The payload includes standard-looking control fields ( IV_DIRECTION , IV_SNDPTR , IV_SNDPRN , IV_RCVPTR , IV_RCVPRN ) and the actual failed-row payload as IT_CSV . IT_CSV is populated via a Logic Apps expression (shown as @{ ...Outputs... } in the screenshot), which is the bridge between the prior transform step and the RFC call. The response side indicates table-like outputs: ET_RETURN and ET_DOCNUMS . 7.1 From CSV to IDocs I’ll cover the details of Destination workflow #2 in Part 2. In this post (Part 1), I focus on the contract and the end-to-end mechanics: what the RFC expects, what it returns, and how the created IDocs show up in the receiving workflow. Before looking at the RFC itself, it helps to understand the payload we’re building inside the IDoc. The screenshot below shows the custom segment definition used by the custom IDoc type. This segment is intentionally shaped to mirror the columns of the CSV input so the mapping stays direct and easy to reason about. Figure: Custom segment ZONLINEORDER000 (segment type ZONLINEORDER ) This segment definition is the contract anchor: it makes the CSV-to-IDoc mapping explicit and stable. Each CSV record becomes one segment instance with the same 14 business fields. That keeps the integration “boringly predictable,” which is exactly what you want when you’re persisting rejected records for later processing. The figure below shows the full loop for persisting failed rows as IDocs. The source workflow calls the custom RFC and sends the invalid CSV rows as XML. SAP converts each row into the custom segment and creates outbound IDocs. Those outbound IDocs are then received by Destination workflow #2, which processes them asynchronously (one workflow instance per IDoc) and appends results into shared storage for reporting. This pattern deliberately separates concerns: the first destination workflow identifies invalid rows and decides whether to persist them, SAP encapsulates the mechanics of IDoc creation behind a stable RFC interface, and a second destination workflow processes those IDocs asynchronously (one per IDoc), which is closer to how IDoc-driven integrations typically operate in production. Destination workflow #2 is included here to show the end-to-end contract and the “receipt” side of the connector scenario: Triggered by the SAP built-in trigger and checks FunctionName = IDOC_INBOUND_ASYNCHRONOUS extracts DOCNUM from the IDoc control record ( EDI_DC40 / DOCNUM ) reconstructs a CSV payload from the IDoc data segment (the fields shown match the segment definition) appends a “verification info” line to shared storage for reporting The implementation details of that workflow (including why it is invoked from the agent tooling) are covered in Part 2. 7.2 Z_CREATE_ONLINEORDER_IDOC - Contract overview The full source code for Z_CREATE_ONLINEORDER_IDOC is included in the supporting material. It’s too long to reproduce inline, so this post focuses on the contract—the part you need to call the RFC correctly and interpret its results. A quick note on authorship: most of the implementation was generated with Copilot, with manual review and fixes to resolve build errors and align the behavior with the intended integration pattern. The contract is deliberately generic because the goal was to produce an RFC that’s reusable across more than one scenario, rather than tightly coupled to a single workflow. At a high level, the RFC is designed to support: Both inbound and outbound IDoc creation It can either write IDocs to the SAP database (inbound-style persistence) or create/distribute IDocs outbound. Multiple IDoc/message/segment combinations IDoc type ( IDOCTYP ), message type ( MESTYP ), and segment type ( SEGTP ) are configurable so the same RFC can be reused. Explicit partner/port routing control Optional sender/receiver partner/port fields can be supplied when routing matters. Traceability of created artifacts The RFC returns created IDoc numbers so the caller can correlate “these failed rows” to “these IDocs.” Contract: Inputs (import parameters) IV_DIRECTION (default: 'O') — 'I' for inbound write-to-db, 'O' for outbound distribute/dispatch IV_IDOCTYP (default: ZONLINEORDERIDOC ) IV_MESTYP (default: ZONLINEORDER ) IV_SEGTP (default: ZONLINEORDER ) Optional partner/port routing fields: IV_SNDPRT , IV_SNDPRN , IV_RCVPRT , IV_RCVPRN , IV_RCVPOR Tables IT_CSV (structure ZTY_CSV_LINE ) — each row is one CSV line (the “table-of-lines” pattern) ET_RETURN (structure BAPIRET2 ) — success/warning/error messages (per-row and/or aggregate) ET_DOCNUMS (type ZTY_DOCNUM_TT ) — list of created IDoc numbers for correlation/traceability Outputs EV_DOCNUM — a convenience “primary / last created” DOCNUM value returned by the RFC 8. Concluding Remarks Part 1 established a stable SAP ↔ Logic Apps integration baseline: CSV moves end‑to‑end using explicit contracts, and failures are surfaced predictably. The source workflow reads CSV from Blob, wraps rows into the IT_CSV table‑of‑lines payload, calls Z_GET_ORDERS_ANALYSIS , and builds one outcome using two fields from the RFC response: EXCEPTIONMSG and RETURN / MESSAGE . The destination workflow gates requests, validates input, and returns only analysis (or errors) back to SAP while handling invalid rows operationally (notification + optional persistence). On the error path, we covered three concrete patterns to raise workflow failures back into SAP: the default connector exception ( SENDEXCEPTIONTOSAPSERVER ), named exceptions (explicit ABAP contract), and message‑class‑based errors (SAP‑native formatting via FORMAT_MESSAGE ). On the remediation side, we added a realistic enterprise pattern: persist rejected rows as custom IDocs via Z_CREATE_ONLINEORDER_IDOC ( IT_CSV in, ET_RETURN + ET_DOCNUMS out), using the custom segment ZONLINEORDER000 as the schema anchor and enabling downstream receipt in Destination workflow #2 (one run per IDoc, correlated via DOCNUM ). Part 2 is separate because it tackles a different problem: the AI layer. With contracts and error semantics now fixed, Part 2 can focus on the agent/tooling details that tend to iterate—rule retrieval, structured validation outputs, prompt constraints, token/history controls, and how the analysis output is generated and shaped—without muddying the transport story. Appendix 1: Parse XML with schema In this section I consider the CSV payload creation as an example, but parsing XML with schema applies in every place where we get an XML input to process, such as when receiving SAP responses, exceptions, or request/responses from other RFCs. Strong contract The Create_CSV_payload step in the shown implementation uses an xpath() + join() expression to extract LINE values from the incoming XML: join( xpath( xml(triggerBody()?['content']), '/*[local-name()="Z_GET_ORDERS_ANALYSIS"] /*[local-name()="IT_CSV"] /*[local-name()="ZTY_CSV_LINE"] /*[local-name()="LINE"]/text()' ), '\r\n' ) That approach works, but it’s essentially a “weak contract”: it assumes the message shape stays stable and that your XPath continues to match. By contrast, the Parse XML with schema action turns the XML payload into structured data based on an XSD, which gives you a “strong contract” and enables downstream steps to bind to known fields instead of re-parsing XML strings. The figure below compares two equivalent ways to build the CSV payload from the RFC input. On the left is the direct xpath() compose (labeled “weak contract”). On the right is the schema-based approach (labeled “strong contract”), where the workflow parses the request first and then builds the CSV payload by iterating over typed rows. What’s visible in the diagram is the key tradeoff: XPath compose path (left): the workflow creates the CSV payload directly using join(xpath(...), '\r\n') , with the XPath written using local-name() selectors. This is fast to prototype, but the contract is implicit—your workflow “trusts” the XML shape and your XPath accuracy. Parse XML with schema path (right): the workflow inserts a Parse XML with schema step (“ Parse Z GET ORDERS ANALYSIS request ”), initializes variables, loops For each CSV row, and Appends to CSV payload, then performs join(variables('CSVPayload'), '\r\n') . Here, the contract is explicit—your XSD defines what IT_CSV and LINE mean, and downstream steps bind to those fields rather than re-parsing XML. A good rule of thumb is: XPath is great for lightweight extraction, while Parse XML with schema is better when you want contract enforcement and long-term maintainability, especially in enterprise integration / BizTalk migration scenarios where schemas are already part of the integration culture. Implementation details The next figure shows the concrete configuration for Parse XML with schema and how its outputs flow into the “For each CSV row” loop. This is the “strong contract” version of the earlier XPath compose. This screenshot highlights three practical implementation details: The Parse action is schema-backed. In the Parameters pane, the action uses: Content: the incoming XML Response Schema source: LogicApp Schema name: Z_GET_ORDERS_ANALYSIS The code view snippet shows the same idea: type: "XmlParse" with content: " @triggerBody()?['content'] " and schema: { source: "LogicApp", name: "Z_GET_ORDERS_ANALYSIS.xsd" }. The parsed output becomes typed “dynamic content.” The loop input is shown as “ JSON Schema for element 'Z_GET_ORDERS_ANALYSIS: IT_CSV' ”. This is the key benefit: you are no longer scraping strings—you are iterating over a structured collection that was produced by schema-based parsing. The LINE extraction becomes trivial and readable. The “Append to CSV payload” step appends @item()?['LINE'] to the CSVpayload variable (as shown in the code snippet). Then the final Create CSV payload becomes a simple join(variables('CSVPayload'), '\r\n') . This is exactly the kind of “workflow readability” benefit you get once XML parsing is schema-backed. Schema generation The Parse action requires XSD schemas, which can be stored in the Logic App (or via a linked Integration Account). The final figure shows a few practical ways to obtain and manage those XSDs: Generate Schema (SAP connector): a “Generate Schema” action with Operation Type = RFC and an RFC Name field, which is a practical way to bootstrap schema artifacts when you already know the RFC you’re calling. Run Diagnostics / Fetch RFC Metadata: a “Run Diagnostics” action showing Operation type = Fetch RFC Metadata and RFC Name, which is useful to confirm the shape of the RFC interface and reconcile it with your XSD/contract. If you don’t want to rely solely on connector-side schema generation, there are also classic “developer tools” approaches: Infer XSD from a sample XML using .NET’s XmlSchemaInference (good for quick starting points). Generate XSD from an XML instance using xsd.exe (handy when you already have representative sample payloads) or by asking your favorite AI prompt. When to choose XPath vs Parse XML with schema (practical guidance) Generally speaking, choose XPath when… You need a quick extraction and you’re comfortable maintaining a single XPath. You don’t want to manage schema artifacts yet (early prototypes). Choose Parse XML with schema when… You want a stronger, explicit contract (XSD defines what the payload is). You want the designer to expose structured outputs (“JSON Schema for element …”) so downstream steps are readable and less brittle. You expect the message shape to evolve over time and prefer schema-driven changes over XPath surgery. Appendix 2: Using FORMAT_MESSAGE to produce SAP‑native error text When propagating failures from Logic Apps back into SAP (for example via Send exception to SAP server), I want the SAP side to produce a predictable, human‑readable message without forcing callers to parse connector‑specific payloads. ABAP’s FORMAT_MESSAGE is ideal for this because it converts SAP’s message context—message class, message number, and up to four variables—into the final message text that SAP would normally display, but without raising a UI message. What FORMAT_MESSAGE does FORMAT_MESSAGE formats a message defined in SAP’s message catalog ( T100 / maintained via SE91 ) using the values in sy-msgid , sy-msgno , and sy-msgv1 … sy-msgv4 . Conceptually, it answers the question: “Given message class + number + variables, what is the rendered message string?” This is particularly useful after an RFC call fails, where ABAP may have message context available even if the exception itself is not a clean string. Why this matters in an RFC wrapper In the message class–based exception configuration, the workflow can provide message metadata (class/number/type) so that SAP can behave “natively”: ABAP receives a failure ( sy-subrc <> 0), formats the message using FORMAT_MESSAGE , and returns the final text in a field like EXCEPTIONMSG (and/or in BAPIRET2 - MESSAGE ). The result is: consistent wording across systems and environments easier localization (SAP selects language-dependent text) separation of concerns: code supplies variables; message content lives in message maintenance A robust pattern After the RFC call, I use this order of precedence: Use any explicit text already provided (for example via system_failure … MESSAGE exceptionmsg), because it’s already formatted. If that’s empty but SAP message context exists ( sy-msgid / sy-msgno ), call FORMAT_MESSAGE to produce the final string. If neither is available, fall back to a generic message that includes sy-subrc . Here is a compact version of that pattern: DATA: lv_text TYPE string. CALL FUNCTION 'Z_GET_ORDERS_ANALYSIS' DESTINATION dest IMPORTING analysis = analysis TABLES it_csv = it_csv CHANGING return = return EXCEPTIONS sendexceptiontosapserver = 1 system_failure = 2 MESSAGE exceptionmsg communication_failure = 3 MESSAGE exceptionmsg OTHERS = 4. IF sy-subrc <> 0. "Prefer explicit message text if it already exists IF exceptionmsg IS INITIAL. "Otherwise format SAP message context into a string IF sy-msgid IS NOT INITIAL AND sy-msgno IS NOT INITIAL. CALL FUNCTION 'FORMAT_MESSAGE' EXPORTING id = sy-msgid no = sy-msgno v1 = sy-msgv1 v2 = sy-msgv2 v3 = sy-msgv3 v4 = sy-msgv4 IMPORTING msg = lv_text. exceptionmsg = lv_text. ELSE. exceptionmsg = |RFC failed (sy-subrc={ sy-subrc }).|. ENDIF. ENDIF. "Optionally normalize into BAPIRET2 for structured consumption return-type = 'E'. return-message = exceptionmsg. ENDIF. Common gotchas FORMAT_MESSAGE only helps if sy-msgid and sy-msgno are set. If the failure did not originate from an SAP message (or message mapping is disabled), these fields may be empty—so keep a fallback. Message numbers are typically 3-digit strings (e.g., 001, 012), matching how messages are stored in the catalog. FORMAT_MESSAGE formats text; it does not raise or display a message. That makes it safe to use in RFC wrappers and background processing. Bottom line: FORMAT_MESSAGE is a simple tool that helps workflow‑originated failures “land” in SAP as clean, SAP‑native messages—especially when using message classes to standardize and localize error text. References Logic Apps Agentic Workflows with SAP - Part 2: AI Agents Handling Errors in SAP BAPI Transactions | Microsoft Community Hub Access SAP from workflows | Microsoft Learn Create common SAP workflows | Microsoft Learn Generate Schemas for SAP Artifacts via Workflows | Microsoft Learn Parse XML using Schemas in Standard workflows - Azure Logic Apps | Microsoft Learn Announcing XML Parse and Compose for Azure Logic Apps GA Exception Handling | ABAP Keyword Documentation Handling and Propagating Exceptions - ABAP Keyword Documentation SAP .NET Connector 3.1 Overview SAP .NET Connector 3.1 Programming Guide All supporting content for this post may be found in the companion GitHub repository.Simplify buying and accelerate AI adoption through Microsoft Marketplace
February 25, 2026 |10:00 AM PST Join this month's Microsoft Marketplace customer office hour to explore practical AI use cases available through the Marketplace - from prebuilt AI apps and agents to AI-powered solutions that simplify buying and deployment Learn more and sign up to attend this online event or watch the recording after the session. Accelerating AI adoption through Microsoft Marketplace | February 25, 2026 - Microsoft Marketplace CommunityWhat to Expect from the Copilot & AI Sessions at Microsoft 365 Community Conference
AI isn’t a side conversation at the Microsoft 365 Community Conference—it’s at the center of how work is changing. The Copilot, Agents, & Copilot Services Sessions are designed for anyone who wants to move beyond curiosity and into real-world application. This is an opportunity to learn how Copilot works today and how agents extend it. You will also explore how organizations can govern, scale, and operationalize AI across Microsoft 365. Questions these sessions will help answer: How do we move from experimentation to real value? How do we scale AI responsibly? How do agents fit into the way we already work? What skills do teams need next? Business leaders, IT pros, developers, and community practitioners will join sessions to find practical insights into how AI shows up in your daily work, and what it takes to deploy it responsibly and effectively. There will also be a focus on change management, champion programs, and adoption frameworks, because deploying AI isn’t just a technical decision, it’s a cultural one. From Copilot to Agents: The Shift from Assistance to Action One of the biggest themes across the sessions are the evolution from AI as a helper to AI as an active participant in work. If you’re curious about what “agentic AI” actually means in practice, attending these sessions will make it concrete. Join your peers as you learn how Microsoft 365 Copilot is being extended through agents that reason, act, and automate. Learn about agent orchestration across tools like Copilot Studio, SharePoint, Teams, Planner, and Power Platform. Discover new agent patterns including declarative agents, multi-agent configurations, workflows agents, and computer-use agents. In these sessions you’ll explore how agents can: Take action on your behalf and do more than suggest content. Work across apps, data sources, and workflows. Participate alongside humans as part of the team. Real Adoption Stories (Not Just Demos)! Go beyond feature walkthroughs to focus on how organizations are actually adopting Copilot and agents at scale. In these adoption stories you’ll hear: How Microsoft uses Copilot and agents internally as Customer Zero. What adoption looks like across large enterprises, frontline environments, and regulated industries. Lessons learned from early adopters—what worked, what didn’t, and what they’d do differently. Governance, Trust, and Control Are Front and Center AI adoption only works when people trust it—and trust is built through strong governance. Learn how organizations are balancing innovation with oversight and enabling teams to build and use agents while maintaining enterprise-grade guardrails. A significant portion of the Copilot & AI track is dedicated to: Agent lifecycle management. Security, compliance, and data protection. Preventing oversharing and managing risk. Observability and control using tools like Agent 365, Microsoft Purview, and Copilot Control System. This is especially valuable for IT and security leaders who are being asked to “move fast” without compromising standards. Building with Copilot: No-Code, Low-Code, and Pro-Code Paths No matter where you sit on the technical spectrum, there’s a clear path to learning how to build responsibly and effectively. Not everyone builds the same way and organizations need prompt engineering that delivers results. In these sessions you’ll learn how to choose the right agent type for the job, extending Copilot with enterprise data, and designing agents that are production ready—not just impressive in demos. These sessions are tailored to: Business users and makers getting started with Copilot Studio Low-code developers extending Copilot with workflows, connectors, and prompts Pro developers building advanced agents using APIs, MCP servers, Microsoft Graph, SharePoint Embedded, and Azure AI Copilot in the Flow of Everyday Work Rather than abstract AI concepts, you’ll see end-to-end workflows that demonstrate how Copilot helps people save time, reduce manual work, and focus on higher-value outcomes. The emphasis in these sessions is on practical impact, not hype showing how AI is grounded in real work. These sessions will showcase Copilot and agents embedded into: Meetings, chats, and channels. Task and project management. Content creation and knowledge management. Business processes and frontline operations. Why the Copilot and AI track matters If AI is part of your roadmap, or already part of your day, this track will show you how strategy can meet execution. Join us to explore clear mental models for Copilots and agents, see real examples you can apply to your work, and gain a better understanding of what’s now—and what’s coming next. Each year, #M365Con26 is built around one simple idea: bringing our global community together to learn, grow, innovate, and get hands-on with the technologies shaping the next era of work. This year’s conference delivers our most expansive program yet, including: 200+ sessions, workshops, and AMAs, covering Microsoft 365 Copilot, Teams, SharePoint, OneDrive, Copilot Studio, and more. 100+ Microsoft-led sessions, giving you unprecedented access to the people building the apps and AI capabilities you use every day. A keynote lineup featuring Microsoft leaders including Jeff Teper, Charles Lamanna, Vasu Jakkal, Rohan Kumar, Jaime Teevan, and many more. Deep-dive workshops to elevate your skills with real-world scenarios and hands-on learning. Exclusive attendee parties and networking events where you can connect with peers and icons. You’ll also get the chance to meet hundreds of Microsoft executives, engineers, and product leaders—ask questions, share feedback, and help shape the roadmap of the technologies you rely on. Register now, save $150 with code SAVE150 - https://aka.ms/M365ConRegisterAgentic AI security: Prompt injection and manipulation attacks
As AI apps and autonomous agents gain more reasoning and independence, they also open new pathways for adversarial attacks. Join this webinar and hear how the most critical risks are broken down—prompt injection, goal hijacking, and memory poisoning—and how they the impact real AI applications. Learn practical defenses your teams can implement today, including input validation, behavioral detection, and robust architectural patterns that keep agentic systems aligned and secure. Learn more and sign up to attend this webinar or watch the recording after. Agentic manipulation: Prompt injection, goal hijacking & memory poisoning | Microsoft Community HubLearn to maximize your productivity at the proMX Project Operations + AI Summit 2026
As organizations accelerate AI adoption across business applications, mastering how Microsoft Dynamics 365 solutions, Copilot, and agents work together is becoming a strategic priority. Fortunately, businesses no longer need to rely on speculation — they can gain practical insights with fellow industry professionals during a unique two-day event: On April 21-22, 2026, Microsoft and proMX will jointly host the fourth edition of proMX Project Operations Summit at the Microsoft office in Munich, but this time with an AI edge. The summit brings together Dynamics 365 customers and Microsoft and proMX experts to explore how AI is reshaping project delivery, resource management, and operational decision‑making across industries. On day one, participants will discover how Dynamics 365 Project Operations, Copilot, Project Online, proMX 365 PPM, and Contact Center can strategically transform business processes and drive organizational growth. On day two, they can explore the technical side of these solutions. Secure your spot! What to expect from the summit Expert-led, actionable insights Join interactive sessions led by Microsoft and proMX experts to learn practical AI and Dynamics 365 skills you can use right away. Inspiring keynotes Gain future-focused perspectives on Dynamics 365, Copilot, and AI to prepare your organization for what’s next. In between our special guests we have Microsoft's Rupa Mantravadi, Chief Product Officer, Dynamics 365 Project Operations, Rob Nehrbas, Head of AI Business Solutions, Archana Prasad, Worldwide FastTrack Leader for Project Operations, and Mathias Klaas, Partner Development Manager. Hands-on AI workshops Take part in workshops where Sebastian Sieber, Global Technology Director (proMX) and Microsoft MVP will show the newest AI features in Dynamics 365, giving you real-world experience with innovative tools. Connect with industry leaders Engage with experts through Q&A sessions, round tables, and personalized Connect Meetings for tailored guidance on your business needs. Real customer success stories Hear case studies from proMX customers who are already using Dynamics 365 solutions and learn proven strategies for successful digital transformation. Who should attend? This summit is tailored for business and IT decision-makers that are using Dynamics 365 solutions and want to drive more business impact with AI, but also for those who might be planning to move away from other project management solutions such as Project Online and need practical guidance grounded in real-life implementations. Date: Apr 21 & 22, 2026 | 2 -Days event Location: Microsoft Munich, Walter-Gropius Straße 5, Munich, Bavaria, DE, 80807 Ready to maximize your productivity? Register here.proMX Project Operations + AI Summit 2026
Ready to learn how you can turn your project challenges into business success with Dynamics 365 AI solutions? Join Microsoft and proMX on April 21st and 22nd at Microsoft Munich for the in-person proMX Project Operations + AI Summit 2026 – Turning data into productivity assets with Copilot + Agents in Dynamics 365 solutions. On DAY 1, discover how Dynamics 365 Project Operations, Copilot, Project Online, proMX 365 PPM, and Contact Center can strategically transform business processes and drive organizational growth. On DAY 2, we’ll take a deep dive into the technical side of these solutions. ✅Book a one-to-one slot with proMX, proMX customers or our Microsoft guests ✅ Explore our booths to talk to Microsoft MVPs and proMX experts ✅ Exchange ideas, project expertise, and more! Register now to secure your spot at this exclusive and free event: proMX Project Operations + AI Summit 2026 You still have questions? Feel free to contact us! Jelena Yaruchyk (Global Marketing Director) - Jelena.Yaruchyk@promx.net Kateryna Marchak (Marketing and Content Lead) - Kateryna.Marchak@promx.netBuilding an AI solution is just the first step - turning it into revenue is where real growth begins
As partners bring AI-powered apps and agents to market at record speed, many struggle with how to price, license, and operationalize their offers for long-term success. This article breaks down how to design monetization early and avoid common pitfalls when publishing on Microsoft Marketplace. Learn how the Marketplace Monetization Checklist helps SaaS and container offers establish secure licensing, effective pricing models, and scalable fulfillment. See how App Advisor translates guidance into action, and how AI Envisioning sessions help align teams on strategy before you build and publish. If you want Marketplace to become a predictable revenue engine—not just a distribution channel read the full article to learn how to monetize your AI apps and agents and drive sustainable growth. Read more: Monetize your AI apps and agents on Marketplace to realize growthLevel up your Python + AI skills with our complete series
We've just wrapped up our live series on Python + AI, a comprehensive nine-part journey diving deep into how to use generative AI models from Python. The series introduced multiple types of models, including LLMs, embedding models, and vision models. We dug into popular techniques like RAG, tool calling, and structured outputs. We assessed AI quality and safety using automated evaluations and red-teaming. Finally, we developed AI agents using popular Python agents frameworks and explored the new Model Context Protocol (MCP). To help you apply what you've learned, all of our code examples work with GitHub Models, a service that provides free models to every GitHub account holder for experimentation and education. Even if you missed the live series, you can still access all the material using the links below! If you're an instructor, feel free to use the slides and code examples in your own classes. If you're a Spanish speaker, check out the Spanish version of the series. Python + AI: Large Language Models 📺 Watch recording In this session, we explore Large Language Models (LLMs), the models that power ChatGPT and GitHub Copilot. We use Python to interact with LLMs using popular packages like the OpenAI SDK and LangChain. We experiment with prompt engineering and few-shot examples to improve outputs. We also demonstrate how to build a full-stack app powered by LLMs and explain the importance of concurrency and streaming for user-facing AI apps. Slides for this session Code repository with examples: python-openai-demos Python + AI: Vector embeddings 📺 Watch recording In our second session, we dive into a different type of model: the vector embedding model. A vector embedding is a way to encode text or images as an array of floating-point numbers. Vector embeddings enable similarity search across many types of content. In this session, we explore different vector embedding models, such as the OpenAI text-embedding-3 series, through both visualizations and Python code. We compare distance metrics, use quantization to reduce vector size, and experiment with multimodal embedding models. Slides for this session Code repository with examples: vector-embedding-demos Python + AI: Retrieval Augmented Generation 📺 Watch recording In our third session, we explore one of the most popular techniques used with LLMs: Retrieval Augmented Generation. RAG is an approach that provides context to the LLM, enabling it to deliver well-grounded answers for a particular domain. The RAG approach works with many types of data sources, including CSVs, webpages, documents, and databases. In this session, we walk through RAG flows in Python, starting with a simple flow and culminating in a full-stack RAG application based on Azure AI Search. Slides for this session Code repository with examples: python-openai-demos Python + AI: Vision models 📺 Watch recording Our fourth session is all about vision models! Vision models are LLMs that can accept both text and images, such as GPT-4o and GPT-4o mini. You can use these models for image captioning, data extraction, question answering, classification, and more! We use Python to send images to vision models, build a basic chat-with-images app, and create a multimodal search engine. Slides for this session Code repository with examples: openai-chat-vision-quickstart Python + AI: Structured outputs 📺 Watch recording In our fifth session, we discover how to get LLMs to output structured responses that adhere to a schema. In Python, all you need to do is define a Pydantic BaseModel to get validated output that perfectly meets your needs. We focus on the structured outputs mode available in OpenAI models, but you can use similar techniques with other model providers. Our examples demonstrate the many ways you can use structured responses, such as entity extraction, classification, and agentic workflows. Slides for this session Code repository with examples: python-openai-demos Python + AI: Quality and safety 📺 Watch recording This session covers a crucial topic: how to use AI safely and how to evaluate the quality of AI outputs. There are multiple mitigation layers when working with LLMs: the model itself, a safety system on top, the prompting and context, and the application user experience. We focus on Azure tools that make it easier to deploy safe AI systems into production. We demonstrate how to configure the Azure AI Content Safety system when working with Azure AI models and how to handle errors in Python code. Then we use the Azure AI Evaluation SDK to evaluate the safety and quality of output from your LLM. Slides for this session Code repository with examples: ai-quality-safety-demos Python + AI: Tool calling 📺 Watch recording In the final part of the series, we focus on the technologies needed to build AI agents, starting with the foundation: tool calling (also known as function calling). We define tool call specifications using both JSON schema and Python function definitions, then send these definitions to the LLM. We demonstrate how to properly handle tool call responses from LLMs, enable parallel tool calling, and iterate over multiple tool calls. Understanding tool calling is absolutely essential before diving into agents, so don't skip over this foundational session. Slides for this session Code repository with examples: python-openai-demos Python + AI: Agents 📺 Watch recording In the penultimate session, we build AI agents! We use Python AI agent frameworks such as the new agent-framework from Microsoft and the popular LangGraph framework. Our agents start simple and then increase in complexity, demonstrating different architectures such as multiple tools, supervisor patterns, graphs, and human-in-the-loop workflows. Slides for this session Code repository with examples: python-ai-agent-frameworks-demos Python + AI: Model Context Protocol 📺 Watch recording In the final session, we dive into the hottest technology of 2025: MCP (Model Context Protocol). This open protocol makes it easy to extend AI agents and chatbots with custom functionality, making them more powerful and flexible. We demonstrate how to use the Python FastMCP SDK to build an MCP server running locally and consume that server from chatbots like GitHub Copilot. Then we build our own MCP client to consume the server. Finally, we discover how easy it is to connect AI agent frameworks like LangGraph and Microsoft agent-framework to MCP servers. With great power comes great responsibility, so we briefly discuss the security risks that come with MCP, both as a user and as a developer. Slides for this session Code repository with examples: python-mcp-demoBringing AI fluency to every corner of the organization (even yours!)
Ashley Masters Hall joined Microsoft more than five years ago, just a month after earning her undergraduate degree. She’s currently a learning manager at Microsoft, focused on AI skilling for business pros. In this first blog post in a series of three, she shares her insightful and relatable perspective on AI fluency and skills for everyone in the organization. I was driving to an appointment recently, and I was reminded of the days when we used to print out directions at home and then try to follow them while driving. That was until GPS and later map apps came along. Suddenly, we had real-time guidance and rerouting from our phones. At first, we might have been reluctant to give up our familiar (though inefficient and unsafe) habit of wrangling printed directions behind the wheel. But once we experienced the speed and simplicity of GPS, there was no going back. AI is having its GPS moment. Two years ago, AI was this mysterious thing, a shiny object we weren’t sure we needed. Now, it’s the default way to navigate (and even orchestrate) work. Just like GPS didn’t replace driving, AI doesn’t replace thinking. It removes friction, gives us faster paths, and lets us focus on better outcomes. So the question isn’t, “Will AI change work?” It already has. The real question is, “Are you fluent enough to lead with it?” Regardless of your role or team, AI is likely already part of your world. In this blog post, the first in a series of three, I share practical tips, including six easy steps, that turn AI from a buzzword into your work GPS. What I mean by AI fluency (no jargon, I promise) Let me first define AI fluency. AI fluency is the degree of understanding and ability to interact effectively with generative AI. It’s recognizing when AI will add value and inspiration, plus having the skills to incorporate it into your workflows and tasks. AI skills are now foundational for everyone in every department. These skills not only help set you apart but also help keep you in the mix as work and roles evolve. We’ve seen a shift in the job market toward skills efficiency rather than work experience. In fact, according to the January 2025 LinkedIn Economic Graph Work Change Report: AI Is Coming to Work, “By 2030, 70% of the skills used in most jobs will change, with AI emerging as a catalyst.” Five years ago, when I was interviewing for jobs, my differentiators were my work experience and the Microsoft Certifications I had earned, which verified my expertise and abilities with Microsoft Office apps. But in interviews last year, most of the questions I got were about how I use AI today and how I’d apply it in the role. That’s the reality: AI fluency isn’t a future skill, it’s a “now skill.” It’s the differentiator that hiring managers seek. They want to know the real-world ways that you’re putting AI to work. Why this matters for your role (for every role, actually) If you’re wondering what AI looks like in your day-to-day work, you’re not alone. Let’s explore its practical applications for different teams and tasks. Marketing. If you’ve ever stared at a blank page trying to translate “We need a campaign” into an actual brief, AI can get you to a starting point fast. Maybe not the final answer, but a decent first draft that you can shape in your voice, for your audience and your goals. Sales and other customer-facing roles. AI is fantastic for meeting prep and follow-ups, like summarizing account notes, pulling themes from call transcripts (if they’re available), and drafting a clean recap email for you to personalize. The magic isn’t the email itself; it’s being able to more quickly and more clearly join in the conversation. Finance. Sometimes the hard part isn’t the analysis but the explanation. AI can help you draft the narrative: what moved, why, what questions a leader might ask, and what you should verify before you hit Send. Human resources (HR). AI can help turn good intentions into clear language, like job descriptions that match the role, onboarding plans that don’t overwhelm people, and summaries of themes you’re hearing so you can act on them before they get lost in the noise. Operations and program management. If your job involves herding context across multiple stakeholders, AI can help you turn chaos into structure, with action trackers, risk lists, crisp status updates, and decision logs you don’t have to rewrite every time. Legal/compliance (with the right guardrails). AI can help you triage, summarize, and spot inconsistencies. And then (pay close attention to this part) people do the actual review. Fluency includes knowing when to stop and bring a human into the loop. If you’re looking for a practical first step, regardless of your role, start with something familiar: email. Watch How to Prompt: Drafting Emails, and learn how Microsoft 365 Copilot can help you get to a first draft in seconds. One thing we shouldn’t gloss over here: AI can speed you up, but it doesn’t take responsibility for you. If you send it to a customer, put it in a deck, or use it to make a decision, you own it, so don’t forget the human layer. That’s the job. It doesn’t need to be scary, but it shouldn’t be overlooked. Get started with AI this week (an easy six-step plan that fits in your real life) If you do one thing after reading this post, don’t make it “Learn AI.” Make it “Pick one task you already do, and run it through a better workflow.” Choose one repeatable workflow. Think of something you do weekly, like meeting prep, a status update, a customer recap, a brief, or a summary. Decide what “better” means to you. Faster? Clearer? Fewer back-and-forth edits? More consistent output? Start with low-risk inputs. Use public info or your own notes while you get more comfortable with the tools. Give Microsoft 365 Copilot a task. Tell it the goal, context, source, and expectations, including the format you need, like bullets, a table, an email, or a memo. Check the work. Verify the facts, including names, numbers, and anything sensitive. Ask Copilot what might be missing. Save the good prompt. Keep it for future reference and reuse. No need to reinvent the wheel every time. Start strong with AI Skills Navigator When people ask me where to start, I have a simple answer: AI Skills Navigator, an agentic learning space that helps you build AI skills (even without a technical background) by bringing together AI-powered skilling experiences, credentials, and training. It makes learning feel approachable. Flexible formats, like short videos, AI‑generated podcasts, quick summaries, and guided skilling sessions, fit naturally into your day in the ways you learn best. This mix of formats helps you get started without feeling overwhelmed, stay engaged as priorities change, and keep up your momentum. AI doesn’t have to feel big. Make it small. Pick one task. Do it once with help. Keep what worked. Repeat it next week. And if you want an easy way to stay on track, start with (and keep coming back to) AI Skills Navigator. It’s like a learning “home base.” Assess where you are, choose a pathway by role or function, and build a habit of learning with small, steady steps. You may be surprised by how far a little AI fluency can take you.
