We are an OpenAI Select Partner. Here is what it changes and what it does not.

OpenAI named AI Operator an OpenAI Select Partner in the OpenAI Partner Network on 12 September 2026.

Announcement posts are usually the least useful thing a company publishes. So here is the version with the actual content in it.

The formal announcement, the one OpenAI approved, is on our press page.

What the partnership gives us

Three things, in order of how much they matter to a client.

Enablement and certification. OpenAI runs structured certification tracks for partners, across ChatGPT deployment, solutions design, API work and Codex. Our delivery team goes through them. What we teach in the room stops being our reading of the documentation and starts being trained against the source.

Roadmap proximity. Partner product updates arrive before the blog post does. For a client planning a rollout two quarters out, that is the difference between designing for what exists and designing for what is about to exist.

A registration channel. When we help a company deploy and adopt OpenAI at scale, that work can be registered with OpenAI directly. Practically, it means the company we trained is no longer invisible to the vendor whose product they just standardised on.

What it does not change

We are not becoming an OpenAI shop.

We run 15 managed agents in production, and a lot more besides. We have published that work for months. None of it is moving because of a badge.

The reason is not diplomacy. It is that our clients' problems are not vendor-shaped. A private equity firm digitising deal memos and a marketing agency rebuilding its brief process need different models, sometimes in the same week, and a trainer with one allegiance gives them worse advice. We pick the model for the job and we say why.

If a partner badge ever starts bending that advice, the badge is the thing that goes.

The part nobody puts in the announcement

Here is what we actually see. Most AI rollouts do not fail on technology. They fail in week three.

The licences get bought. The training day happens. Everyone leaves impressed. Then the work gets busy, the old process is right there, and by week three the seats are sitting unused while the invoice renews.

That has almost nothing to do with which model a company chose. It is a habit problem wearing a technology costume. It is also why our programme is twelve weeks and not two days, and why participants build tools for their own jobs instead of watching someone demo one.

A partner tier does not fix week three. Daily practice does. The partnership just means the material people practise with comes from closer to the source.

What is next for us

Our delivery team starts the OpenAI certification tracks this month. We are working through which of the eleven badge tracks actually map to what we sell, rather than collecting all of them.

And we are going back to clients whose ChatGPT rollouts we ran, to ask whether they want that work registered with OpenAI. That is their call to make, not ours.

If your rollout is stalling

If your team has ChatGPT seats and usage dropped off after week three, the useful question is not which model to switch to. It is where in the week the new habit is losing to the old one.

Our free AI Audit answers that. It takes a few minutes and it tells you where adoption is breaking.

Run the free AI Audit