Field Notes 03 of the Forward Deployed Engineering series
Proximity turned out to be the technology
Three years ago, AWS ran an experiment: what happens if you put applied scientists and strategists directly in the room with customers, instead of shipping them a deck?
That experiment became the Generative AI Innovation Center. I lead the startup practice there.
The public numbers tell part of the story: over 1,000 customers, and more than 65% of last year’s projects reached production — in an industry where most AI pilots quietly die.
But the number I care about is the pattern behind it. Every engagement that made it to production had the same shape: our team embedded with theirs, a business problem their leadership actually agreed on, and short cycles that forced decisions instead of deferring them.
The engagements that struggled? Almost always the mirror image. A technical spec without a business owner. Feedback loops measured in weeks. Security and legal discovered at the end instead of invited at the start.
Proximity turned out to be the technology.
So when AWS announced a $1B commitment to Forward Deployed Engineering last week, I didn’t read it as something new. I read it as a lesson we’ve been learning for three years, finally at full scale.