OpenAI execs say companies need to do 3 things right to get employees using AI
Summary
OpenAI’s product and engineering leads, Olivier Godement and Sherwin Wu, told the BG2 podcast that organisations need a pragmatic, results-focused approach to get employees using AI. They highlighted three essentials: secure top-level buy-in while empowering a focused “tiger team,” create clear benchmarks (“evals”) driven by real operator knowledge, and continually monitor and iterate on progress. The execs stressed that much of the value in enterprises lives in employees’ heads, so cross-functional teams and bottom-up input are crucial to produce useful AI tools.
Key Points
- Combine top-down leadership support with a small, empowered “tiger team” that blends technical skills and subject-matter expertise.
- Define clear benchmarks or “evals” to measure AI performance — these often need bottom-up input from frontline staff.
- Monitor those benchmarks closely and iterate; progress can be part art, part science and may require model fine-tuning.
- Give teams permission to experiment: start small, learn, then scale successful pilots.
- Organisational knowledge is frequently tacit, so including operators in design and evaluation is vital for useful AI rollouts.
Author take
Punchy and practical: this is a short playbook for any organisation trying to move from AI curiosity to actual usage. Leadership buy-in, a sharp cross-functional team, and measurable goals — get those three right and you stop playing with demos and start shipping impact.
Why should I read this?
If you’re responsible for rolling out AI at work, this is the no-nonsense checklist you want. It saves you time by cutting through vendor hype and showing the real levers that move projects from experiments to everyday tools. Seriously — stop guessing and use a tiger team + evals approach.
