Amazon Web Services is putting $1 billion into a new engineering unit built to sit inside its customers’ own teams. The group, called Forward Deployed Engineering, will send small pods of AWS staff to work face to face with clients on their toughest AI problems.

Francessca Vasquez, AWS vice president of frontier AI engineering and services, said the company already had pieces of this capability scattered across the business. Now it is pulling everything into one unit with a shared playbook for how AI gets deployed.

What the new AWS unit actually does

Each engagement starts small. AWS sends about five or six engineers to a company for a stretch of around 45 days. They sit with the client’s own staff, work through specific technical blockers, and help wire AI into existing workflows rather than starting from scratch.

The company plans to hire thousands of these Forward Deployed Engineers over time, all funded through the initial $1 billion commitment. Early customers already working with AWS teams include the NFL, the NBA, the Allen Institute, Cox Automotive, Ricoh, and Southwest Airlines.

A pricing model built around results, not hours

AWS is charging fixed, outcome-based fees for these engagements instead of billing by the hour. That is a break from how consulting giants like Tata Consultancy Services and Infosys typically price similar work, and it puts pressure on AWS to show measurable gains fast.

The forward deployed engineer title itself is not new. Palantir Technologies popularized the role more than a decade ago, embedding engineers inside defense and enterprise clients to get software working in the real world rather than in a lab.

AWS joins a crowded race for enterprise AI talent

OpenAI and Anthropic have both been building out their own forward deployed teams this year, betting that enterprise AI adoption stalls without hands-on help. Google is reportedly hiring thousands of similar engineers of its own.

For AWS, the bet is that whoever helps a company get its first AI project working keeps that company’s cloud spending for years after. Vasquez framed it simply: getting everybody under one roof, with a common way of measuring deployment success.

The unit has already started work with its first clients, well before the thousands of planned hires are in place.