IBM Just Bet Its Consulting Arm on OpenAI. The Model Is the Easy Part. Getting It Into Your Business Is the Business.

On 13 August, IBM announced a strategic partnership with OpenAI to deploy AI at scale across enterprises' core operations, embedding OpenAI's frontier models, GPT-5.6, Codex and ChatGPT Work, into IBM Consulting Advantage, the platform IBM uses to deliver consulting. The two will build industry solutions for financial services, government, telecom and retail, and IBM will train and certify tens of thousands of consultants on OpenAI's technology.

What the deal actually is

Strip it down and this is a distribution deal. OpenAI has the models; IBM has the thing OpenAI does not, tens of thousands of consultants inside the world's largest enterprises and the trust to be let near their core systems. IBM gets to sell frontier AI as an outcome rather than an API, and OpenAI gets its models embedded where enterprise budgets actually live. The cybersecurity angle, IBM joining OpenAI's Daybreak program and pairing frontier models with its Autonomous Security service, is the same idea in a different suit: capability plus delivery.

Read where the value sits

Here is the part worth sitting with. The model is not the scarce thing in this deal. GPT-5.6 is available to anyone with a credit card. What IBM is monetising, and what it is training an army to do, is the last mile: turning a general capability into agentic workflows that run a specific company's finance, procurement and operations. That is not a model problem, it is an integration and decision problem, and it is where the hours, the trust and the margin are. When the most enterprise-savvy company on earth bets on the delivery rather than the model, that tells you where the real work is.

Two ways to buy the last mile

There are, broadly, two ways an enterprise gets that last mile. One is the route IBM just supersized: services, a consultancy that assembles the models, the integration and the change management into a bespoke engagement, billed by the hour. The other is product: a packaged app that already encodes the decision for a specific job, installs on your stack, and runs. Both deliver outcomes; they are different businesses. The services route is flexible, expensive, and rebuilt each time. The product route is narrower, repeatable, and cheaper to own, which is the logic behind a companion app that optimises one decision on top of your ERP rather than a project that reinvents it.

What to take from it

For anyone deploying AI in their business, the IBM and OpenAI deal is a useful mirror. It confirms that the base models are becoming interchangeable, the same signal as tiered model pricing and software you can now build from a prompt. And it confirms the value has moved to the last mile of making that capability decide and act inside your specific operation. The only real question is whether you buy that last mile as a project you commission or a product you install.