OpenAI just published a guide distilling interviews with executives at Philips, BBVA, Mirakl, Scout24, JetBrains, and Scania on how they're scaling AI. The findings don't read like a vendor success story — they read like a warning to anyone still treating AI deployment as a technical rollout.
"Scaling AI is less about 'rolling out AI' and more about building the conditions where people trust it, adopt it, and improve it over time."
Five patterns came up consistently. They're worth sitting with.
The 5 patterns
1. Culture before tooling — The fastest adoption came from building literacy, confidence, and permission to experiment — not from shipping the platform first.
2. Governance as an enabler — Where security, legal, compliance, and IT were brought in early as design partners, teams moved faster later. Fewer reversals. More trust. The orgs that treated governance as a blocker paid for it.
3. Ownership over consumption — AI scaled when teams could redesign their own workflows and build with AI — not just consume it as a feature someone else configured.
4. Quality before scale — The organisations that earned lasting trust defined what "good" meant early, invested in evaluation, and were willing to delay launches. The ones that didn't created backlash they had to unwind.
5. Protecting judgment work — The most durable gains came from hybrid workflows: AI lifting the ceiling on expert reasoning and review, not replacing it. Volume gains without judgment protection don't hold.
The real lesson
None of this is new advice in isolation. The interesting part is that it's converging across very different industries and company sizes — a bank, a healthcare company, a dev tools shop, an automotive group.
The common thread: the companies pulling ahead aren't simply moving faster. They're treating AI as an operating layer and a leadership discipline, not a feature release.
That means workflow design, not just model selection. Governance built in from day one, not bolted on after the first compliance scare. And teams that own their AI surface area rather than waiting for IT to hand them a pre-built tool.
What to do
- Running an AI pilot that keeps stalling? Check culture first. If people don't have permission to experiment and fail safely, the tooling won't save you.
- Governance conversations still happening in security reviews post-launch? Pull them earlier — ideally into the design phase. It's counterintuitive but it speeds things up.
- Teams consuming AI features but not building with them? That's a ceiling. The organisations going furthest are the ones where teams can redesign their own workflows.
- Tempted to scale before quality is solid? Don't. The organisations in this guide that earned trust moved more slowly at first, then compounded.
Worth downloading the full guide if you're responsible for an AI rollout — it includes a one-page leadership diagnostic and a checklist for pressure-testing readiness.
Source: How enterprises are scaling AI — OpenAI | Frontiers of AI Executive Guide (PDF)
✏️ Drafted with KewBot (AI), edited and approved by Drew.
Top comments (0)