Agency17 June 20263 min read
Getting AI into production at a scale-up: the in-between problem
Thirty engineers, a codebase that works, everyone now using AI, and delivery somehow isn't faster. Scale-ups hit a bind the startup and enterprise playbooks don't fix. Here's the one that fits.
The short answer
Scale-ups have the hardest time getting AI into production because they're in between. The codebase is real enough that lax standards hurt, unlike a startup's greenfield, but the team is too lean to run an enterprise's year-long governance program. Worse, AI usually speeds up typing while the real bottleneck moves to review and release, so the team feels faster and delivers slower. The path that fits is incremental and on your own stack: fix the delivery system, not just the keyboard, win one team, and keep your standards intact.
Short version: picture thirty engineers, a codebase that works and that people depend on, and a quarter where everyone started "using AI." Delivery isn't faster. It might be slower, and review is backing up. This is the scale-up bind. You're past the size where you can just let everyone hack like a five-person startup, and you're short of the budget and slack an enterprise spends on a year-long program. The fix isn't either of those playbooks. It's an in-between one, and it starts by fixing the right bottleneck.