Most of the early attention on artificial intelligence focused on generative AI, but the conversation has shifted recently to include agentic AI, or AI systems that can act autonomously. We think of these agents as AI assistants designed to automate and execute business processes—alongside users or on their behalf. In 2024, Microsoft introduced agents in Microsoft 365 Copilot, and early this year we announced that we’re putting agents at people’s fingertips in Microsoft 365 Copilot Chat. That makes it easier than ever for employees to quickly and securely adopt agentic AI.
For enterprises, this is an opportunity to develop line-of-business (LOB) agents that can transform business processes and lead to ROI from adoption of Microsoft 365 Copilot across the organization. Using Microsoft resources, like the Microsoft 365 Dev Center, enterprise developers can create agents to fulfill industry needs and custom use cases, from simple agents that retrieve and reason over business system information to advanced ones that execute business processes autonomously.
Dow, the global materials science company, is pioneering the development of these LOB agents. In this post, I’ll go deeper into how they’re using Copilot and agents to reimagine their shipping operations by developing an agentic workflow that will save them millions of dollars.
Dow’s ambitious data readiness project set the stage for Microsoft 365 Copilot
Data readiness is crucial for maximizing the value of AI deployments, and enterprises often take steps to improve data quality, reliability, and accessibility in advance. Dow exemplifies this approach, as they have prioritized harnessing company data as a strategic asset in recent years. As part of that effort, they rolled out their Integrated Data Hub, which helps improve data access and discoverability across business lines and empowers teams with streamlined analytics.
Dow creates massive amounts of system-level data, and by connecting the information in the Integrated Data Hub with Copilot, employees can get insights much faster than they could with traditional reporting and analytics. Connecting Copilot and agents to organizational knowledge sources using connectors in Microsoft Copilot Studio also helps enterprises get more reliable, relevant responses from Copilot and enhances the capabilities of agents.
Dow also understood that it was critical to the company’s AI initiatives to train employees on how to interact with data and AI. As part of the Integrated Data Hub project, they rolled out a data and analytics literacy program, which teaches employees how to manage data, make better decisions with it, build data visualizations, and use AI tools.
Shipping invoices provide a high-value use case for Dow to deploy Copilot and agents
Next, Dow needed to identify the business processes that would benefit most from the increased efficiency or speed delivered by applying Copilot and agents.
They held a joint workshop with Microsoft to explore where LOB agents would be effective at resolving pain points, focused on freight operations. Dow oversees up to 4,000 daily outbound shipments across all modes of transport, and the logistics team was manually checking invoices for misapplied fees and errors that could result in overpayment. They were unable to efficiently audit them because of the volume and variety of formats, including 100,000 PDFs annually—some up to 20 pages long. It quickly became clear that the process was ripe for assistance from AI.
Dow ultimately decided to start with North American road-based freight, which reduced the variability and complexity compared to international shipments and other modes of transport. By narrowing the focus, they created a manageable subset of invoices to prove the value of Copilot quickly—another important tactic for enterprise adoption.
"With an annual freight spend topping several billion dollars, Copilot and the agent help identify overcharges to improve efficiency and free teams to focus on higher value work,” said Jeff Tazelaar, Dow’s Global ISC Innovation Director.
Copilot Studio streamlined development of Dow’s autonomous agent
With their target identified, Dow reimagined manual processes with an agentic workflow in freight invoicing. They decided to develop two LOB agents, and they included an expert from the logistics team in the development process to maintain focus on the project’s business goals and to reduce friction with adoption among logistics employees once they were ready to scale.
Using Copilot Studio, where enterprise developers can build, deploy, and manage LOB agents, Dow created an autonomous agent that monitors incoming email for attached PDF invoices and structures data for analysis. It reviews freight rates, accrued costs, and invoice rates to ensure they’re aligned with negotiated terms. Then the agent color codes discrepancies, noncompliant invoices, and billing inaccuracies and surfaces the data in a dashboard for employees.
"Copilot Studio has been a game-changer,” Tazelaar said. “Insights actually show up in the decision-making process.”
Dow believes employees will be more likely to adopt the agent precisely because it surfaces during the decision-making process. Once the autonomous agent completes its work, employees turn to the Freight Agent, a prompt-and-response agent. The Freight Agent was trained on 43,000 invoices from the Integrated Data Hub, and it enables employees to use natural language to investigate inaccuracies and compare data among carriers, truckloads, product types, and more.
Only a few weeks into a proof-of-concept pilot, Dow employees and agents were already collaborating to flag anomalies among thousands of invoices and highlight potential savings. Overcharges and invoice errors that would have taken days or weeks to identify can be uncovered in a matter of seconds. Employees can also see duplicate invoices or track elevated fees for unloading delays—or any other data point that might reveal mistakes or inefficiencies.
Dow’s results underscore two key pieces of guidance for any successful deployment:
- Ground Copilot in organizational data to improve its capabilities and responses.
- Design agentic workflows that free employees from time-consuming tasks while keeping them in the loop to make strategic decisions.
A bright future at Dow with Copilot and line-of-business agents
Dow anticipates saving millions of dollars in freight costs once the agentic system is scaled to shipping modes globally. And that’s just scratching the surface. They’ve already identified over 100 use cases for logistics and operations across their supply chain. For example, they plan to leverage a network of autonomous agents that could relay data about weather, potential delays, and changes to cost and then compare deviations to the original contract terms so that discrepancies are identified before shipments even arrive.
Their AI initiative extends beyond shipping and logistics, as Dow has expanded Copilot to nearly 60% of the company. They’re also developing agents for other business groups, like one that helps the research and development team with patenting new innovations.
"This is just the beginning as we continue to transform our supply chain and business processes with Copilot, driving both operational efficiency and sustainability," said Dow CIO Melanie Kalmar.
As the agentic era of AI takes shape, we couldn’t be more excited to work alongside enterprises like Dow to discover what else is possible.
Learn more about Microsoft 365 Copilot and agents
For more information about Copilot and the ways Dow is using LOB agents to transform its business processes, check out the links below.
- Get started building agents by visiting the Microsoft 365 Dev Center for decision guides, code samples, documentation, and tutorials.
- Build agents with Microsoft Copilot Studio and Microsoft Visual Studio.
- Learn how to deploy agents on our instructions page.
- For more information about how Dow is using Copilot and agents, read our WorkLab report and the case study.
About the Author
Srini Raghavan is CVP of Microsoft 365 Copilot Ecosystem in the Experiences & Devices division. He is focused on building a premier developer platform and agent ecosystem for Microsoft 365 Copilot, collaborating with strategic ISVs and enterprises globally. Additionally, he oversees the Microsoft Teams platform ecosystem, encompassing Collaborative Apps, core platform capabilities, Calling, Meetings, and Copilot in Teams.
With over 19 years of leadership at Microsoft, Srini has held pivotal roles across engineering and product management, shaping foundational products such as Microsoft Teams, Office 365, Office Communications, Lync, and Skype for Business. As a founding member of Office 365, he played an instrumental role in launching its first cloud-based service, while driving service engineering and product management for Office Live Meeting, Skype for Business, and Microsoft Teams.
Microsoft