Had a conversation with a fintech CTO recently. A year ago - 40 developers. Now - 10. Not layoffs to cut costs. A deliberate rebuild around AI.
“We kept the people who know how to think, not just code.”
I’m hearing this more and more. So let me break down what it actually means for teams - from a PM’s perspective.
What’s happening
AI is shifting the balance between “thinking” and “doing”. Developers are still needed. Just different ones. Or the same ones - with different skills.
Who stays
A pattern I’ve assembled from conversations:
Who stays:
- Systems thinkers - people who see the full architecture
- Those who can slice work into clean, precise prompts
- People with business context - they understand the why, not just the how
- Detective-style debuggers - can navigate AI-generated code
Who leaves:
- Coders without systems thinking
- Those who don’t adapt
- Specialists on tasks where AI is simply faster
Role boundaries are blurring.
What changes for PMs
My process over the past year:
Before: PRD by hand → hand off to dev → wait → test → iterate → document. Everything manual, kept in my head.
After: PRD with AI → sometimes build a prototype with AI in a day and show the team alongside the PRD → iterate (still needed) → document with AI. My head cleared up a lot. I don’t have to keep the entire project context in memory - I just need to know where to store it, in what form, and how to use it. AI became a companion that took over a ton of routine, amplified results, and made room for things that always mattered but never had time to happen.
Practical steps
1. Audit your tasks
Go through your backlog:
- AI handles it better (routine, generation, first drafts)
- AI assists, human controls (reviews, architecture)
- Human only (strategy, negotiation, creative)
2. Think roles, not people
Not “fire 30 people”. But “what roles do we actually need”. Sometimes one senior with AI equals three juniors. Sometimes not - depends on the work.
3. Give people time to learn
Those 10 out of 40 weren’t just the best. They got time to learn a new way of working. That’s an investment.
4. New metrics
Old: lines of code, closed tickets. New: quality of decisions, time to value.
What people fear
“AI will replace everyone” - no. AI will replace those who don’t learn to work with it. Excel didn’t replace accountants - it replaced people doing math on paper.
“It’s urgent” - the urgency is overstated. But you can’t ignore it either.
“Quality will drop” - it might, if you just swap people for AI without restructuring your processes.
Bottom line
AI transformation isn’t about headcount cuts. It’s about redesigning how products get built.
Start with yourself. Try AI on your own tasks. Figure out where it helps. Then scale.
Those 10 out of 40 are the pioneers. Their roles are more interesting now.
Sources
- Interview with a fintech CTO (anonymous, 2026)
- Claude Code Documentation
- Personal experience transforming PM processes