The pitch used to be so clean you could fit it on a napkin. Build a model smart enough, and the value takes care of itself.

On June 14, OpenAI quietly retired that idea. It launched the OpenAI Partner Network — a $150 million program to train and certify 300,000 consultants by the end of 2026 — and tucked inside the announcement was one sentence that should stop anyone who spent the last three years believing the model was the magic.

“The limiting factor for seeing value from AI in the enterprise is no longer model capabilities.”

Read it twice. The company that sells the models just told you the models aren't the problem.

The founding partners are not startups. They are Accenture, Bain, BCG, McKinsey, and PwC — the exact firms that have spent decades billing the Fortune 500 for slide decks about transformation. OpenAI sorted them into three tiers: Select, Advanced, and Elite, ranked on sales performance, co-selling activity, and how many deployments a partner can show. There are planned specializations in Codex, cybersecurity, and agents.

Strip the branding and you have a sales channel. A very expensive, very credentialed sales channel.

The job description for these 300,000 people is revealing. OpenAI says partners will help customers “identify use cases, redesign workflows, connect OpenAI products with existing systems, and support workforce adoption.” Two of those phrases matter more than the other two. “Redesign workflows” and “support workforce adoption” are consultant dialect. Translated into plain English, they mean: figure out which steps a human still does, and figure out how to get the rest of the humans to stop resisting.

That is the real product here. Not the model. The permission structure around firing people, delivered by a logo your board already trusts.

My Opinion

I'll be blunt. This is the most honest thing OpenAI has said in a year, and it came out sideways.

For three years the entire industry sold the dream that intelligence was the scarce resource — that once the benchmark went up, the economy would reorganize itself. GPT-5.5 ships, Claude Opus 4.8 ships, Gemini 3.5 ships, and the benchmarks keep climbing. Yet here is OpenAI spending $150 million to admit that none of that is what's slowing enterprises down. The bottleneck was never the IQ of the software. It was that real companies are messy, and getting value out of AI requires someone to sit in a conference room and rewire how work actually happens. That job pays $400 an hour and it has McKinsey's name on the invoice.

Here's what bugs me. The same consulting firms that will now bill enterprises to “adopt” OpenAI are the ones who profit on both ends — the deployment fee on the way in, the restructuring fee when the headcount comes out. OpenAI gets a 300,000-person army it doesn't have to pay salaries for. The consultancies get a shiny new SKU. And the worker whose “workflow” just got “redesigned” gets a calendar invite titled “org update.” Everybody in that chain wins except the person the chain is built on.

So watch the next twelve months closely. If model capability were really the engine of AI value, OpenAI would be pouring $150 million into research, not into certifying consultants. It chose the consultants. That tells you exactly where the company thinks the money is now — and it isn't in the lab.

Author: Yahor Kamarou (Mark) / www.humai.blog / 18 Jun 2026