Don’t go off vibes: data on how AI is changing software development | Featuring Span Field CTO Stephen Poletto

Every engineering leader is hearing the same story right now: AI is making teams dramatically faster. More output, more leverage, more shipped. But we wanted to separate hype from reality, so we invited Span Field CTO Stephen Poletto to come and share what he’s seeing from the real-world data.

One moment from this conversation cut through the noise. An engineering leader at a public B2B company said they were deep in annual performance reviews, comparing net impact delivered in 2025 vs. 2024 — and despite all the new AI tooling, it didn’t feel remarkably different. That quiet disconnect between headlines and lived experience set the tone for the entire discussion.

If you’d been there, here’s what you’d still be thinking about:

The easiest metrics to reach for are the ones most likely to mislead us: When leaders try to measure AI-driven productivity, they default to what’s easy: lines of code, commits, pull requests, story points. But these metrics often optimize activity, not impact. AI can inflate output without improving outcomes. It can create more code, smaller PRs, faster estimate, while quietly increasing complexity and maintenance burden. The result: teams look busier on dashboards, but customers don’t necessarily feel the difference. And the business doesn’t move.

Writing code fast was never the bottleneck: AI is exceptionally good at generating code, but that speed doesn’t mean the entire process of writing and shipping code is faster in all. The constraints now show up in code review, quality guardrails, integration, rollout, and adoption. As one visual Stephen referenced made clear: the bottleneck has always lived at the edges of the development lifecycle, not in typing code. AI didn’t remove it. It exposed it.

Velocity is up — and so is rework: Benchmark data from the past year shows teams shipping more pull requests as AI adoption increases. But it also shows more rework, more review cycles, and more churn after deployment. Faster output doesn’t automatically translate to more value. The hard question leaders now have to answer is no longer “are we shipping faster?” — it’s “are customers actually experiencing more impact?”

For complete event notes and the recording, apply to join Enrich.