At the end of June, AWS announced a $1 billion investment in a Forward Deployed Engineering organization: thousands of AWS engineers embedded directly inside customer teams to build and ship agentic AI systems. Two days later, Microsoft introduced its Frontier Company, built explicitly on the same forward-deployed model. The two biggest clouds made the same bet in the same week, and the bet was not on a better model. It was on people who show up and build.
We have been that embedded engineering team for mid-market companies for years, so we read these announcements with some interest. They confirm something we argue constantly, and they leave a gap our clients live inside.
What the hyperscalers just conceded
For three years the industry debate has been about capability: which model, which platform, which benchmark. The FDE investments are a quiet admission that capability was not the constraint. If adoption were blocked on model quality, you would fix it by shipping a better model. You do not spend a billion dollars embedding engineers with customers unless the real blocker is everything between the model and a production system: the data, the integration, the governance, the verification, the people.
That gap between what the technology can do and what an organization can actually get out of it is the whole story of enterprise AI right now. The hyperscalers just priced it, publicly.
Who the embedded motion is built for
Read AWS's announcement closely and the target market is explicit: organizations in regulated industries, financial services, and government, with launch customers like the NFL, the NBA, and Southwest Airlines. That is not a criticism. Embedding a team of hyperscaler engineers is expensive, and the economics only work on accounts large enough to repay it.
But it draws a line, and most mid-market companies are below it. The companies with the messiest estates, the thinnest platform teams, and the most to gain from real implementation help are exactly the ones no hyperscaler will build a team around.
What the model gets right
Strip away the scale and the FDE model makes three claims worth keeping, because they match what we see work in the field.
Embedded beats advisory. The engineers sit inside your team, in your environment, shipping against your systems. Assessments and roadmaps do not ship systems; builders do.
Outcomes beat deliverables. The engagement is structured around a business result, not a stack of documents. If nothing runs in production, nothing happened.
Self-sufficiency is the exit. A good engagement leaves your own team able to operate and extend what was built. If the partner has to stay forever, that is a dependency, not a capability.
None of those three requires a hyperscaler. They require senior engineers, a delivery practice built for embedding, and an incentive to make themselves unnecessary.
The mid-market version of the choice
If the hyperscaler will not build a team around you, the realistic options are a commodity AI shop that hands you a demo, a large integrator running the same big-logo economics as the clouds, or a small senior team that works embedded. Agentic delivery has quietly changed that last option's math: a handful of senior engineers using agents well now covers ground that used to take a much larger bench, which is exactly what makes the embedded model affordable below the enterprise line.
Whoever you pick, hold them to the same three properties the hyperscalers are now selling: embedded, outcome-scoped, and designed to leave. If you are trying to figure out whether your organization is ready for that kind of engagement, our AI readiness assessment is where we usually start, and our AI implementation framework is what the embedded work runs on.