For most of the last three years, OpenAI has argued for speed. Move fast, ship the frontier model, let the market catch up. Then, on 11 September 2026, in a companywide meeting reported by Bloomberg, chief executive Sam Altman told staff the company is open to pacing its frontier AI development, and said he hopes other major labs will do the same. OpenAI has separately asked US lawmakers whether coordinating a slowdown across companies would even be legal under antitrust rules.

This is a reversal. The company that built its identity on rapid, largely unconstrained development is now asking whether the industry, itself included, should ease off.

Worth sitting with, if you run an agency.

Agency owners rarely say this out loud, but a lot of AI anxiety in this sector isn't really about the technology. It's about pace. A competitor mentions an AI tool at a networking event, a client asks why something hasn't been automated yet, a supplier's newsletter announces another model release, and the quiet worry sets in: everyone else must be moving faster than us.

That worry has always been a little unfair on agency owners, most of whom are running lean teams with no dedicated AI function and a full client roster to service. But it's harder to sustain even in theory now. If the company that has spent years telling everyone else to hurry up is privately wondering whether the industry's own pace is sustainable, then sprinting to keep up with a moving target that its own inventor is reconsidering starts to look less like diligence and more like noise.

I saw the alternative work in practice earlier this year, running the AI Accelerator training for The Turner Agency, a 36-person events and film business. I didn't try to get the whole team fluent in a single day. The programme ran across six sessions, each one building on the last, with room in between for people to actually use what they'd learned before the next idea landed. It wasn't the fastest way to look busy on AI. It was the way that got people using it properly, reflected in 4.6 out of 5 satisfaction across 131 survey responses, without a single dissatisfied response logged.

That's the pattern worth borrowing from OpenAI's own hesitation. Pacing yourself deliberately isn't the same as falling behind. It's a decision about how change actually sticks in a business made of people, not just tools.

Having spent years inside AI transformation programmes at organisations including Microsoft and Vodafone, the pattern holds at every scale: the businesses that get the most durable value from AI are rarely the ones moving fastest. They're the ones who worked out what mattered to their own operation and their own clients, then built capability at a pace their people could actually absorb.

None of this means agencies should ignore AI, or wait for permission before starting. The point cuts the other way. It removes a bad reason for rushing. If competitors seem to be racing ahead, that's worth checking rather than assuming, because the businesses that look fastest from the outside are often the ones with the least to show for it a year later. And if the industry's own frontrunner is now asking Congress whether slowing down is even allowed, an agency choosing a considered pace over a sprint isn't behind. It's doing exactly what the moment calls for.

The AI Accelerator Diagnostic exists for agencies who want to work out that pace deliberately, rather than by accident: what's actually worth doing, in what order, over what timeframe is realistic for the team doing it. A half day on site, a structured look across six dimensions, and a written roadmap within a week.

Go wisely.

Source: Bloomberg, reported 11 September 2026, via coverage in TechCrunch, PYMNTS and Vantage Markets