If you've sat in enough platform team planning sessions lately, you've probably seen the same pattern emerge: an engineering manager splits the stack intentionally—a free AI coding tool for exploratory spikes and experiments, a paid or self-hosted solution for the production monorepo where security and compliance actually matter. It sounds reasonable on the whiteboard. Three weeks later, nobody's touching the expensive path at all.

Why the Split Fails in Practice

The problem isn't that teams make bad decisions during planning—it's that they stop making decisions entirely after rollout. Nobody mapped repos to paths. There's no feedback loop telling you whether engineers are actually using the paid tool where it matters, or just defaulting to whatever's fastest at 2 PM on a Thursday when they're blocked on a refactor.

The 7-Field Fit Score Framework

The author proposes evaluating each repository against seven fields before assigning an AI coding path: codebase sensitivity (public vs internal), team size and velocity requirements, security posture needs, context window demands for the domain, code review culture fit, cost sensitivity at that layer of the stack, and operational overhead tolerance. The goal is forcing explicit mapping rather than letting parallel tool adoption happen by accident.

What This Means for Platform Teams

This isn't really about AI coding tools—it's a generic platform team problem wearing new clothes. Any time you let multiple paths coexist without clear ownership and measurement, drift happens. The free tier gets used everywhere because friction is low. The paid tier becomes the thing people feel guilty about not using.

Making Multi-Path Strategy Actually Work

If you're running or considering two AI coding paths, treat it like any other infrastructure decision: define success criteria per repo type upfront, instrument usage from day one, and build in a regular review cadence. Don't let the split be a planning artifact that nobody ever validates against reality.

Key Takeaways

  • Map repos to tools before rollout, not after adoption drift becomes visible
  • Treat AI coding tool strategy like any other platform infrastructure decision
  • Instrument usage immediately—don't discover low adoption three sprints later
  • Define clear criteria for when a repo belongs on free vs paid paths

The Bottom Line

The dual-path model can absolutely work, but only if your platform team treats it as an active architecture decision rather than a one-time setup. Otherwise you're just paying for two tools and using one.