AWS commits $1bn to embed AI engineers with customers

TL;DR:

• Amazon Web Services is creating a division of “forward-deployed engineers” who embed directly with customers to help them adopt AI, backed by an initial $1bn.
• The plan is to send pods of engineers to clients for 45-day stints, writing production code and driving agentic AI into their workflows.
• AWS wants “thousands” of staff in the unit even as Amazon has cut more than 30,000 corporate jobs since October.

Amazon is betting that the barrier to enterprise AI is not the models but the last mile of putting them to work. Its cloud arm has committed an initial $1bn to a unit of forward-deployed engineers — versatile staff who sit alongside customers, navigate their internal politics and write production-grade code to make AI deliver.

A rare hiring bright spot

The model is not new. Palantir has run forward-deployed engineering for over a decade, and Salesforce, Anthropic and Google Cloud offer their own versions; AWS concedes it is “a bit late”. But demand is surging: one report cited a 42-fold rise in demand for such roles between 2023 and 2025, making this a rare area of tech hiring even as Amazon sheds corporate headcount elsewhere.

AWS frames success as speed — getting customers to a working product or new skill faster than traditional project work allows. Initial clients include the NBA and electronics firm Ricoh, with the unit unveiled at a two-day Washington customer event expected to yield further government-cloud announcements.

The move speaks directly to the adoption gap troubling UK boards. Fresh research this week estimated that UK businesses waste £67bn a year on failed AI projects, with weak execution — not ambition — the culprit. Hyperscalers selling hands-on delivery, rather than just capacity, is a direct response to that failure rate, and UK enterprises weighing large AI programmes are an obvious target market.

Looking forward

Embedding engineers is expensive and hard to scale, which is why AWS is pairing it with a headline number rather than a headcount. If the pods measurably shorten the path from pilot to payoff, expect rivals to raise their own commitments — and expect “we will send our engineers to you” to become a standard line in enterprise AI sales.