When it comes to enterprise AI adoption, few companies have moved as aggressively as EY. Under Global Head of AI, John Thompson, the firm built and deployed a multi-million dollar custom AI environment to their 420,000 global employees.

But despite their massive AI investment, John doesn’t believe in chasing traditional AI ROI metrics. In fact, he didn’t even try to measure AI ROI in the first year.

And when other leaders who haven’t rolled out AI yet come to him with fears of falling behind, he tells them not to sweat it.

Here’s why he says you shouldn’t be in a rush to make AI investments or prove the financial value of them.

Why you need to slow down rather than speed up

John has had conversations with leaders from hundreds of companies who have all said something like: “If we don’t have AI up and running next week, we’re out of business”.

What he tells them is: You see your competitors talking about AI so you think you’re falling behind, but actually they’re probably still figuring it out. This is a good time for you to start, too.

“Over the next two to three to five years, if you don’t get involved, yes, you’ll lose your edge in the marketplace,” John says. “But it’s not a heart attack kind of thing. It’s more like a slow erosion kind of thing."

The trap that these leaders are susceptible to is rushing into poor AI decisions that definitely won’t lead to ROI vs. taking longer on more a strategic approach.

"You’re not in a bad place because you’re not using it. You’re not in a great place, but you do have time."

He explains that unlike previous tech hype cycles (like the metaverse and blockchain), AI has a clear trajectory toward long-term adoption.

So instead of scrambling and launching a series of expensive mistakes, businesses should take a slower and more structured approach to their AI adoption. Here’s what he recommends

1. Be realistic about your limitations. John wanted his team to be able to do whatever they wanted in their AI platform without worrying about data privacy, so he built a custom AI environment.
   
   It’s more secure and was ultimately cheaper than rolling out existing enterprise solutions (read: single digit millions vs. double digit millions). So if you’re going to put up a ton of guardrails or limit AI use cases, don’t just throw a ChatGPT Enterprise plan at your team and wonder why they don’t use it – consider making a larger but more strategic long-term investment.
   

2. Define your core use cases before picking an LLM. Not all AI models are the same, and many have vastly different strengths. So before you invest in the hyped up chatbot of the moment, make sure it can solve real business problems for you. Start by asking what you really want to do with it (code generation, document review, etc.) and then find an LLM that excels at that.
   
   
3. Talk about your AI investments. A lot. John didn’t need to convince anyone to try out the AI environment he built. By the time it launched, people were reaching out to him. Share your AI plans with your team regularly so by the time it’s ready for them to work with, they’re excited to try it.

The real danger isn’t failing overnight—it’s becoming irrelevant over time. So the key here is finding a pace between delaying indefinitely and making hasty investments out of fear.

Why proving ROI is not as urgent as you think

We already established that EY’s AI investment was sizable, so the natural next question is how the returns on that investment are shaping up. But for John, ROI is more of an inevitability than an ongoing metric.

"I understand that you have to make financial sense out of things,” he said. “It’s not just all puppies and merry-go-rounds. But AI is changing the world, and it’s going to continue to change the world."

EY paid for the AI system centrally in the first year – meaning the org covered the inference costs, individual users and teams could play around with it, and no one was required to justify the cost of their AI usage.

"We knew that our 420,000 people would be better if they understood Gen AI. So in the first year, we just built it and let people use it. We didn’t really measure anything."

In essence, they allowed organic adoption to drive value – and it did. Some of their internal AI use cases became client-facing products and services. The team’s experiments have turned into monetizable offerings that EY sells to clients.

Now, AI usage is paid for on the team or business unit level. Instead of tracking org-wide ROI metrics for AI that may not apply to everyone, EY has let each pocket of the company make the right investments for their individual projects and ideas.

His advice to other leaders looking for AI ROI:

1. Let go of the traditional metrics. AI’s value is often realized in indirect ways, like productizing your own internal workflows. So be open to determining the value of AI after adopting it instead of before.
   
   
2. Realize that no one knows how to do this yet. “I haven’t seen anybody in the world who’s really good at understanding the streams of possible value and the potential ability to capture that value,” John said. Learn from other people’s successes, but don’t look for a plug-and-play strategy.
   
   
3. Don’t nickel and dime. Part of realizing the full potential of AI is just accepting that some aspects of it will be immeasurable. Your team will be more efficient, your company will save time, all of this will translate into cost savings. But if you’re worried about proving all that value down the last cent, you’re going to waste a lot of time.

The bottom line: Don’t get rigid on what ROI has to look like or wait to adopt AI until you have a solid strategy for measuring its value. Just adopt it and start finding value.

It’s not too late — but it’s time to get started

"I've had conversations with people who have said to me, ‘If we just wait it out, it'll go away’. And I'm like, no, it’s not like a rat infestation," John says.

So take the time to set your team up for long-term success, focus less on initiatives for 12-24 month ROI, and be open to new value streams appearing where you didn’t expect them.

And if you liked John’s insights, catch him live in conversation with our CEO, Greg Shove.