The last mile

An MIT study found ~95% of enterprise AI pilots never produce measurable P&L impact. The models were rarely the problem. The last mile was.

Last week, AWS committed $1 billion to closing that gap — Forward Deployed Engineering: thousands of engineers embedded directly inside customer teams, co-building production agentic AI systems.

I’ve spent the last three years at the AWS Generative AI Innovation Center working with startups and enterprises on exactly this problem. The pattern is remarkably consistent: pilots don’t stall on model quality. They stall on data access, security reviews, legacy integration, and workflows nobody redesigned.

You can’t fix any of that from a distance. You fix it standing next to the people who own the workflow — with security, legal, and ops in the room from day one.

That’s what forward deployment means. Not a bigger services bench. Engineers who sit inside your team, build with AI agents rather than just talking about them, and stay until the system is running in production.

Congratulations to Francessca Vasquez and the team on the launch. The ambition is right: move enterprise AI from “promising pilot” to “running in production” — in days, not months.

The real question for enterprise AI in 2026 isn’t which model you pick. It’s who’s standing next to you when it hits production.