Why EY built an entire AI platform for enterprise-scale agentic AI

EY had just successfully trained its 400,000 employees on generative AI and deployed the technology throughout the organization when agentic AI came along to disrupt workflows yet again. Now the Big Four accounting and IT consultancy, officially known as Ernst & Young Global Limited, needed a comprehensive plan for deploying AI agents that added business value while operating within the compliance requirements of EY's highly regulated industry. Employees would need a new round of training to add agentic tools to their AI arsenals and learn to work alongside agents.

The solution was a single, multilayered AI "operating system" built in collaboration with Microsoft and Nvidia that sits atop the generative AI foundation and unifies intelligence, orchestration, data, workflows, governance and domain experience, according to an EY case study.

In this episode of Enterprise Apps Unpacked, Julie Teigland, EY's global vice chair of alliances and ecosystems, explains how the platform works and how other companies can follow a similar strategy.

Defragmenting AI development

EY leadership was well aware of the challenges other adopters of agentic AI encountered when trying to move beyond pilots and scale the technology across the enterprise. The company had identified AI as central to its ambitious goal of $100 billion in annual revenue, so solving the scale problem would be critical.

Fragmentation was already becoming a problem. "We had some really valuable AI capabilities, data assets and use cases coming across the organization, but to be honest, they weren't connected," Teigland said. "Lots of different teams were experimenting with different models. We had data across multiple systems, and governance approaches were kind of evolving. Scaling that became increasingly complex."

She likens the platform to a wedding cake with three layers, with data as the foundational layer, orchestration and workflow in the middle, and an intelligence layer at the top that contains an AI model catalog. "AI really only scales when all three come together: intelligence, data and governance," she said.

The platform has also driven innovation, with the company developing more than 50,000 agents over nine months, according to Teigland. And in line with past IT initiatives, EY "drinks its own champagne" before sharing what it has learned with clients. She said one client used EY's agentic platform approach in its AI-infused SAP S/4HANA transformation.

Such an ambitious undertaking might seem beyond the capabilities of smaller companies, but Teigland said that's not the case, citing the example of a shipbuilding company that hired EY to build a similar layered platform to improve collaboration between humanoid robots and employees. "SMBs need the same type of model, the same layer wedding cake approach in terms of getting their data right and making sure they've got that governance baked in," she said.

Other topics discussed in the podcast include the following:

• The business benefits the platform delivers to EY, its clients and partners.
• The respective roles of EY, Microsoft and Nvidia in the development partnership.
• How the agentic AI platform is used internally.
• The biggest lessons learned.

David Essex is an industry editor who creates in-depth content on enterprise applications, emerging technology and market trends for several Informa TechTarget websites.