The report's data reveals a fundamental tension. 38% of organizations now ship at least twice as much as they did before AI. Only 8% can point to a return worth writing home about.

The data and insights are clear. Platform engineers are the ones who close that gap. This report sets out how, addressing the following:

What did all that extra throughput actually earn? Shipping more is not the same as earning more. We break down where the return is landing, where it is going negative once licenses, tokens and review overhead are counted, and why a fifth of the industry cannot answer the ROI question at all

What is actually standing in the way? We asked teams to rank what most obstructs scaling AI. The answer at the top of that list is not the model, it’s the platform.

What are platform teams being asked to own now? Central model connectivity, agent runtimes, inference workloads, and a customer list that has grown to include data teams, business users and the agents themselves. The report covers how the role is expected to evolve, what that means for the engineers doing it, and why cheap generation makes the platform team more valuable rather than less.

What are the leading organizations doing differently? The thing to build now has a name. The Agentic Engineering Platform is what the internal developer platform becomes when agents are users of it rather than tools bolted onto the side of it. We set out the four capabilities that only pay off as a set, the reason the return arrives at one specific step of the maturity ladder rather than accumulating gradually, and what it costs to get stuck one rung below it.

Everyone in this industry now has access to the same models. What they build around them is the only thing left to compete on. This report is about what that looks like when it works.