What does it look like when one of the world’s largest professional services firms builds an AI agent whose only job is to teach 400,000 employees to build their own AI agents?
Welcome to Episode 10 of The Business AI Playbook, the podcast where I sit down with leaders who are rethinking how organisations develop and enable their people. Each episode goes deep on the real decisions, trade-offs, and results behind AI adoption in the enterprise. If you want to see what production-ready AI training video looks like in practice, try Colossyan free — no credit card required.
Listen to the full episode on Spotify, YouTube, or Apple Podcasts.
EY built an agent whose job is to teach its workforce to build agents
Simon Brown is the global Learning and Development leader at EY. His team is responsible for AI skills across 400,000 employees in 150 countries. Two and a half years into the role, he’s overseen one of the largest corporate AI learning rollouts publicly described to date, and the latest iteration is, in his own words, a world first.
EY just launched an AI agent. The agent’s only job is to teach an EY employee how to build their own AI agent. A short tutorial introduces the user, and then the agent takes over. It asks what would be a good use case for them personally, walks them through the build step by step, and by the end of the session the user has shipped a functional agent they can keep using. About 20,000 people have gone through it in the first few weeks. It’s rolling out to all 400,000.
What I found most striking was how Simon framed the journey. This wasn’t a single project. It was the third stage of a three-stage evolution that mirrors how the technology itself has matured. Stage one was “AI Now,” a voluntary module to introduce people to what AI even is. 350,000 people took it. That’s 90% of the organisation, the most popular training EY has ever run. Stage two was “Gen AI as Thought Partner,” which had over 500,000 completions across two modules. Stage three is the agent-building agent. Each one was built when the technology was ready and the workforce was ready, not before.
“This is rolling out to 400,000 people. An agent to teach people how to build agents. By the end of the training, you’ve actually hands-on built your own agent.”
The data behind the strategy is the part that should make every L&D leader pause. EY’s own Work Reimagined survey, across about 1,500 organisations, found that 88% of people now use AI regularly. Only 5% use it at an advanced level. The 5% group is unlocking around 1.5 days of productivity per week. They got there by putting in 80+ hours of dedicated learning. The gap between “I use ChatGPT sometimes” and “I run my workday with AI” is not a vibe gap. It’s a learning gap, measured in hours.
In this episode, we discussed
- Why EY built an agent to teach agent-building, instead of a course about agent-building
- The three-stage AI learning rollout: AI Now (350k completions, 90% uptake), Gen AI as Thought Partner (500k+ completions), and the agent-building agent
- The Work Reimagined data: 88% of workers use AI, only 5% are advanced, advanced users get 1.5 days back per week
- Why hitting 400,000 employees with anything takes 6 to 7 months, not “a day” like a demo
- The 50,000 agents already in use across EY and the governance tiers behind them
- How EY tops Coursera’s global benchmark for advanced and intermediate AI skills
- Simon’s personal 30-agent setup, including “the agitator” he uses to challenge his thinking
- Why he vibe-coded his own home automation with Claude, including lights, blinds, and the front door
“Within a day someone could create a simple version. But that’s not going to work at scale across 400,000 people, with all of those use cases, with all of the tracking.”
The bigger lesson here
The most interesting part of this conversation wasn’t the agent itself. It was the principle behind it. EY’s L&D function moved in lockstep with the technology. Awareness when AI was new. Thought-partner usage when chat became the default interface. Agent-building when agents became the next frontier. Most companies are still rolling out “Intro to ChatGPT” courses in 2026. EY is teaching its workforce to build agents.
Simon also said something every senior leader should sit with. A useful agent for yourself is a one-day project. A useful agent for 400,000 people across 150 countries is a six to seven month project. The difference is everything that doesn’t show up in a LinkedIn demo. Data privacy. Ethical evaluation. Security review. Load balancing. Multi-model selection. Response-time optimisation. The gap between demoware and production AI is where most enterprise AI ambitions actually die. It’s also where the next decade of corporate competitive advantage will be earned.
One last thing worth flagging. Simon personally runs 30 AI agents, vibe-coded his home automation with Claude, and builds apps with his teenage daughter on weekends. His principle is that what you play with at home unlocks what’s possible at work. If you’re a senior L&D leader who hasn’t built a personal agent setup yet, that’s worth sitting with too.
How Paramount cut training video production from weeks to hours
Paramount replaced manual video production with Colossyan's AI avatars, helping its global L&D team create training content faster without traditional production bottlenecks.
Read the full story →The full episode is now available on Spotify, Apple Podcasts and YouTube.
About the guest
Simon Brown is the Global Learning and Development Leader at EY (Ernst & Young), responsible for AI skills development across 400,000 employees in 150 countries. He previously served as Chief Learning Officer at Novartis and Bupa. Simon is the co-author of The Curious Advantage, a book on curiosity as the load-bearing skill of the modern workforce, and he hosts the Curious Advantage podcast, where he interviews leaders on curiosity, learning, and the future of work. Connect with him on LinkedIn.
