What happened
On Thursday, Microsoft announced the Microsoft Frontier Company: a $2.5 billion initiative that embeds more than 6,000 industry, engineering, and AI professionals inside customer organizations to design, build, and run AI systems on-site. It will be led by Rodrigo Kede Lima, most recently president of Microsoft Asia, with early engagements at the London Stock Exchange Group, Unilever, Land O'Lakes, and Accenture.
If this sounds familiar, it should — Amazon committed $1 billion to the same play two days earlier, the story in yesterday's issue. In one 72-hour stretch, the two largest cloud companies on Earth committed $3.5 billion to the same admission: AI doesn't produce results until someone installs it inside the business.
The detail almost everyone will miss
Start with what this "company" isn't. It is not a separate legal entity, most of its 6,000 people already work at Microsoft, and the company wouldn't say whether the $2.5 billion is new spending or repurposed budget. Strip the branding and this is a reorganization of Microsoft's existing consulting muscle around one job: deployment.
What's new is the scoreboard — all four AI giants now run deployment businesses, and that only happens when the real money moves. OpenAI stood up a standalone deployment company in May with over $4 billion in backing. Anthropic joined Goldman Sachs and Blackstone on a $1.5 billion services venture the same month. Models are getting cheaper and more interchangeable by the quarter; the profit pool is shifting from selling intelligence to installing it.
Microsoft's own CEO framed the stakes in June: no business should end up "ceding value to a few models that eat everything they see." His test for whether you still control your future — can you swap one AI model for another without losing the system you built around it?
Why this matters if you run a business
Look at the client lists. Stock exchanges, global consumer brands, the Fortune 500. This entire $10-billion-plus deployment race — Microsoft, Amazon, OpenAI, Anthropic — is aimed at the largest companies in the world. Nobody is forward-deploying engineers into a 15-person operation.
The most sophisticated buyers of AI just repriced hands-on deployment as the scarce asset — and none of them are selling it at your size. The diagnosis still applies at every size: the gap between the subscription you pay for and the results you don't see is a labor problem, not a software problem. The giants can now buy that labor from Microsoft. Everyone else has to be deliberate about where they get it.
What to do about it
The useful move is to steal the contract terms, not the headline. The deals announced this week reveal exactly what the smartest buyers now demand from anyone who installs AI — and every one of them fits on a one-page agreement:
- Ownership survives the engagement. Microsoft is pitching "your data and your institutional knowledge stay yours" as its differentiator. Make that a requirement of anyone you hire, not a bonus.
- Payment follows outcomes, not hours. Microsoft calls its new organization "outcome-driven" — a deliberate break from consulting's billable-hour model. Tie any AI engagement to a measured workflow result.
- The system outlives the model. Apply the swap test before you sign: if you changed AI providers next year, would your workflows, data, and training survive the move?
That swap test is the single best question a business can ask before signing any AI engagement — and it came from the CEO selling the engagement. When the vendor tells you how to protect yourself from vendors, take notes.