Selling AI and deploying it inside a customer's walls have turned out to be two different businesses, and the companies that make the models and platforms are no longer willing to leave the second one to someone else.
In an eleven-week span this spring and summer, a run of the most consequential companies in enterprise technology each concluded, independently, that converting AI into revenue meant launching a services organization built around embedded engineers. Salesforce launched its certified partner network in mid-April. Google Cloud committed $750 million to the same idea a week later, at Cloud Next. Anthropic announced a services company with Blackstone, Hellman & Friedman, and Goldman Sachs in early May. OpenAI followed within the week, folding an acquired consultancy into a $4 billion venture led by TPG. Databricks formalized its own delivery organization in June. AWS put $1 billion of its own balance sheet behind a dedicated unit at the end of the month. Microsoft answered two days later, naming 6,000 people and $2.5 billion to a new operating business it insists is not merely another forward-deployed engineering program, even though it functions as one. Snowflake has quietly built the same capability without a flagship announcement at all.
The consistency across companies with otherwise very different business models matters more than any individual figure. A model lab, a cloud platform, a data platform, and a CRM vendor arrived at the same structural answer: whatever they are selling, whether tokens, compute, storage, or workflow, the customer cannot turn it into a working system without someone doing the integration work inside their walls (https://alten.capital/blog/the-new-agentic-fde). That work has historically been left to consulting firms and systems integrators. The vendors building it in-house now say something specific about where they believe the value and the risk actually sit.
It also reveals that "services business" is doing double duty as a term, since these moves are not really the same kind of thing. AWS, Microsoft, Databricks, and Snowflake put their own engineers directly inside customer accounts, funded off their own balance sheets or capitalized as a new operating business that keeps the customer relationship in-house. Anthropic and OpenAI did something closer to spinning out a standalone consultancy: majority-owned by the parent, but capitalized by outside private equity and banks, with a mandate to build their own client base. Google Cloud took a third path, declining to build a customer-facing arm at all and instead paying to embed its engineers inside the systems integrators that already hold those relationships. Salesforce runs a version of both at once, a small internal team built quietly over the past year, sitting underneath a much larger, newly announced certified-partner network that does the actual scaling.
That distinction, who owns the customer relationship and whose balance sheet funds delivery, is worth tracking, because it implies different economics. A vendor that keeps engineers in-house is betting the delivery layer is defensible enough to justify the fixed cost, and stands to capture services margin directly if the bet pays off (https://alten.capital/blog/the-agentic-margin). A vendor financing partner-embedded engineers is betting the opposite: that owning the platform and model is sufficient, and delivery is better rented from firms that already carry the account relationships and the balance sheet for headcount-heavy work. Neither is obviously correct, and each depends on whether the embedded engineering actually produces enough switching-cost lock-in, wherever it sits, to justify the spend.
What is not in dispute is the diagnosis every one of these companies shares: the model is no longer the product, the deployed system is (From Time to Outcomes), and enterprise AI pilots stall on delivery capacity rather than model capability. Every company that makes something enterprises want to run AI on top of is now, in some form, also in the business of putting engineers on a plane. What varies, and what will show up in the numbers over the next several years, is how much of that business each is willing to own.
Alten Capital invests in technology services businesses. Reach out to explore potential partnerships.