How to scale AI: A look inside EY’s journey with Microsoft 365 Copilot and agents

We are in a time of transition. While experimentation with AI continues, many businesses are moving to scale it throughout their organization, driven by the quick evolution of AI agents. By 2028, businesses will employ an estimated 1.3 billion agents, according to an IDC study.¹

Microsoft is streamlining this push to build agents with the resources in the Microsoft 365 Dev Center and the development options in Microsoft Copilot Studio. From January through March of this year, customers created more than 1 million custom agents with Copilot Studio and Microsoft SharePoint, up 130% quarter over quarter.² Soon, this shift toward agentic work will restructure the way businesses run, giving rise to a Frontier Firm that relies on human-led, AI-operated systems.

In this post, I’ll showcase how EY, one of the world’s largest professional services organizations, is leading the way toward that vision with AI and agents. EY has developed numerous agents for Microsoft 365 Copilot, both to boost employee productivity and to help clients transform their businesses. I’ll explore the ways they’ve used Copilot Studio and Microsoft 365 Copilot Tuning in development, as well as the people-centric way they approach deployment. I’ll also highlight their early plans for orchestrating entire agentic teams.

EY accelerates the AI journey with Microsoft 365 Copilot and agents

AI is a critical area of focus for EY, and they have gone all in on Microsoft 365 Copilot as an accelerator for their AI initiatives. The company has:

• Deployed Copilot to 100,000 EY users globally
• Developed an app to help other firms pilot and scale Copilot
• Built AI agents using Copilot Studio, for both internal and external use
• Used Copilot Tuning to train a large language model (LLM) on EY tax expertise
• Helped clients speed up deployment of Copilot with their insights and agents

EY began with their own global rollout. A deployment of that scale involves challenges with data security and dealing with data sovereignty in different jurisdictions across the world, which they addressed in collaboration with Microsoft.

"There was a lot of complexity around making sure we maintain client data security, and how to leverage Copilot in a safe and secure way for our use internally and with clients,” said Paul Clark, EY’s Global Alliance Leader for the Microsoft partnership. “Microsoft really understood that."

Now, EY users operate securely across jurisdictions while using a common AI framework, and leaders are confident they can speed up the next Copilot deployment to all 400,000 employees. They also used their experience to develop Everyday AI with Microsoft 365 Copilot, an app that helps enterprise decision-makers pilot, scale, and reap the benefits of AI responsibly.

Though some lessons are technical, much of what EY learned relates to people. For example, they quickly realized it was crucial to deliver a consistent change management message to employees and to highlight that adopting AI would involve stepping stones rather than a one-and-done effort.

EY emphasized that AI tools amplify abilities rather than replace employees. They showcased the productivity benefits of Copilot, like summarizing meetings and drafting emails, to encourage adoption. Next, they focused on how agents can unlock new use cases and transform workflows to add capacity to each employee.

EY also provides iterative training on Copilot and agents to encourage employees to increase their use and shift their mindset about how AI can transform the way they work. By leveraging EY’s unifying AI platform, EY.ai, their conversational AI assistant, EYQ, and Copilot, EY’s people have achieved a 15% boost in productivity—time that’s now being reinvested into client delivery, continuous learning, and operational excellence.

Microsoft Copilot Studio and Microsoft 365 Copilot Tuning help EY build agents that create value

When EY employees began using Copilot Studio, a comprehensive platform for creating, managing, and deploying agents, it provided a breakthrough in shifting their mentality. Copilot Studio has pro-code options for developers and the ability to integrate with additional, industry-specific foundation models in Azure AI Foundry to create more advanced use cases. But it also features no-code and low-code development for end users and makers, and creating agents in that easy-to-use environment opened employees’ minds to what else they could accomplish.

An agent developed by EY

For example, EY teams have used Copilot Studio to develop the EY PortalOne agent, an AI-powered tax and legal research agent that provides instant access to over 21 million documents, including EY publications, legislation, and templates. This generative AI solution streamlines research for EY tax and legal professionals, empowering them to deliver outcomes with enhanced insight and intelligence.

EY is also gaining value from the company’s long-term investment in the Microsoft ecosystem. The EY Strategies and Transactions team previously created a Microsoft Teams app, SaT Workbench, that facilitates collaboration and administration for engagements. The SaT Workbench team is working with Microsoft to develop a related agent that automates compliance checks within project-specific Workbench sites. This intelligent assistant will help ensure all required documentation and processes are completed before project archival—enhancing efficiency and accuracy in compliance management and improving project oversight.

In addition, EY is an early adopter of the new Copilot Tuning capability in Copilot Studio, which was rolled out in June for organizations in the early access program. Copilot Tuning allows users to train models on organizational data, workflows, and processes to help with domain-specific tasks. Once training is complete, agents can tap into those models to perform tasks with a high degree of accuracy.

EY is using Copilot Tuning capabilities to integrate a tax-domain fine-tuned LLM with Microsoft 365 Graph and their tax expertise. The resulting rich contextualization will further enhance the quality of their tax advisory services while significantly advancing tax and legal research. 

EY sets the stage for multi-agent orchestration and fast-tracks clients’ AI deployments

With a growing number of agents across business lines, EY is looking to the third stage in their journey toward AI maturity: orchestrating cohorts of agents. In public preview in Copilot Studio, multi-agent orchestration enables agents to exchange data, collaborate on tasks, and divide work based on each agent’s capabilities. Though still early in the process, EY is already planning the framework for how they will approach orchestration.

EY’s leaders envision an orchestration agent that will connect people to the right agents and tools to accomplish their work. They’re also planning to deploy a training agent to ensure that other agents are trained consistently and on the right models. Then they anticipate developing an agent to oversee other agents and ensure compliance. Of course, a person will manage the team of agents—what we often call an “agent boss”—to keep employees in the loop while freeing them to focus on high-value work. In effect, the Microsoft platform is fully capable of delivering end-to-end agents that work together.

EY will benefit from the robust AI foundation they’ve created, and their early adoption of Copilot and agents also provides experience for helping clients accelerate their AI journey. They can better advise companies on data sovereignty issues and guide them on how to ensure their data estate is ready for AI—with clean, consistent, accessible data.

They can also offer best practices for change management messaging that encourages not just adoption, but a deeper, more strategic approach. They know from experience that the restructuring required to become a Frontier Firm will demand a whole new way of thinking.

"The journey with AI is not a linear path. It's not like you can simply say, 'I want to change the process,'” said Clark. “You have to completely reimagine the process, and that's where AI agents can really help and shift our understanding of what’s possible."

Learn more about Microsoft 365 Copilot and agents

For more information about Copilot and how EY is using agents to accelerate business, follow the links below. 

• Get started building agents in the Microsoft 365 Dev Center with decision guides, code samples, documentation, and tutorials.
• Build agents with Microsoft Copilot Studio and Microsoft Visual Studio.
• Check out how Microsoft 365 Copilot Tuning can add skills to the models that power agents.
• Learn how to deploy agents on our instructions page.
• Find out more about how EY and Microsoft are collaborating to help enterprises implement Copilot and realize the full potential of AI.
• Download the Everyday AI with Microsoft Copilot app to jumpstart the AI journey.

The views reflected in this article are the views of the author and do not necessarily reflect the views of the global EY organization or its member firms.

¹IDC Info Snapshot, sponsored by Microsoft, 1.3 Billion AI Agents by 2028, #US53361825 and May 2025.

²Microsoft Fiscal Year 2025 Third Quarter Earnings Conference Call, April 30, 2025