Decisions

Two AI Giants Just Became Your Integrator

OpenAI's $4bn DeployCo and Microsoft's $2.5bn Frontier embed vendor engineers in your operations. Name who owns the integration before you sign.

DecisionsBy Servola Tech Desk2026-07-113 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Two AI Giants Just Became Your Integrator

The contract on the operations director's desk

A European operations director sits with an AI deployment contract in front of her. The vendor's proposal is not just a model licence. It includes the vendor's own engineers, on site, to build the deployment, wire it into her logistics systems and keep it running. The pitch is speed, and it is genuine. The quieter question she has to answer is: what is she actually buying?

Two of the largest AI vendors have each stood up a forward-deployed engineering arm to make exactly this offer. OpenAI Deployment Company, nicknamed DeployCo, launched as a majority-owned subsidiary with more than $4 billion of backing from a coalition including TPG, Bain Capital, Brookfield, Goldman Sachs, McKinsey and Capgemini. It stations OpenAI Forward Deployed Engineers inside large enterprises in healthcare, logistics, manufacturing and financial services, and has bought the applied-AI consultancy Tomoro, roughly 150 FDEs, with a further firm called Northslope reported on July 8. Microsoft answered on July 2, 2026 with Frontier Company, a $2.5 billion commitment and about 6,000 industry and engineering experts built on the same embedded model.

The build-vs-buy decision grows a third option

For a decade the enterprise choice was build or buy. Build the capability with your own people, or buy a finished product and integrate it yourself. The forward-deployed model inserts a third path: let the vendor run it. The company that sells you the model now also sells you the engineers who wire it into your operations, and the two arrive as one signature.

The trap: that third option is the most comfortable to sign and the hardest to leave. When the vendor's engineers hold the knowledge of how your workflows are wired, the lock-in moves up a layer. It is no longer the model you depend on, it is the operating model they built inside you. Switching models becomes cheap next to switching the people who understand your own processes.

Name who owns the integration before you sign

The move: before the signature, name in writing who owns the integration knowledge. Require that the runbooks, the architecture decisions and the integration IP are documented and handed to your team as the work proceeds, not held in the vendor's heads. Put an exit clause on the operating model itself: if the engineers leave, your people can run what was built without a new contract.

Ask the second question too: who employs the person at the whiteboard next year. If the answer is always the vendor, you are not buying a deployment, you are renting your own operating model back from the firm that sold you the model. Speed is worth paying for. Control of how your business runs is not worth trading for it.

Frequently asked questions

Is forward-deployed engineering just consulting with a new name?

Not quite. Traditional integrators are model-agnostic and leave you the documentation. Here the integrator is the model vendor, and the embedded engineers deepen your dependence on one platform while they work. The service is real, but the incentive runs toward lock-in.

Does this only matter for very large enterprises?

No. The named targets are large firms in healthcare, logistics, manufacturing and financial services, but mid-sized operators face the sharper version, because they have fewer internal engineers to absorb the handover and are more likely to let the vendor keep running it.

What single clause matters most?

Integration-IP ownership with a runbook handover. If your team can operate and modify the deployment after the vendor's engineers leave, you kept control. If not, every renewal is negotiated from weakness.

The vendors are becoming the systems integrators, and the offer is good enough that most buyers will take it without reading what it does to their leverage. Sign it if the speed is worth it, but sign it with the integration knowledge in your own hands. Own how your business runs, or you will rent it back every year.

OpenAIMicrosoftforward deployed engineersAI deploymentvendor lock-inbuild vs buyenterprise AIprocurement

Apple Says OpenAI Hired Away Its Hardware Secrets

Apple sued OpenAI on 10 July 2026 over unreleased hardware secrets. The channels it names are ordinary hiring, and the fix is offboarding discipline.

4 min read

Artificial Intelligence

OpenAI Just Halved the Price of Everyday AI

GPT-5.6 shipped July 9 in three tiers. Terra matches last year's GPT-5.5 at half the price, and Luna sets a new floor. The owner move is to reprice and route by task tier.

3 min read

Digital Sovereignty

The EU Council Revived the Rule That Scans Private Messages

The EU Council revived the lapsed rule that lets providers scan private messages with AI. For owners, confidential communication is now a governance choice.

2 min read

← All articles