There are two AI conversations happening right now, and they’re not the same conversation.
In engineering rooms, the talk is concrete. Which models are worth using. Where Copilot saves time and where it wastes it. Whether the new agent flows are actually production-ready or still demo-grade. Real tradeoffs, real frustration, real wins.
In boardrooms, the talk is forecast. AI will reshape our business by Q3. We’re integrating AI across every workflow. Expect 40% productivity gains. Big numbers, big confidence, big timelines.
The gap between those two rooms is widening, and that’s the actual story.
What’s actually shipping
In my work, AI is genuinely useful for narrow, well-defined tasks. Generating boilerplate. Translating between languages. Writing first-draft tests. Explaining unfamiliar code. None of this is small — these are real productivity wins for individual engineers.
But the gains are local. A developer who saves 20 minutes on a function isn’t shipping features 20 minutes earlier. Code review is the same. Testing is the same. Architecture decisions are the same. Coordination overhead is the same. The bottleneck moves, but the total project timeline doesn’t shrink the way leadership thinks it will.
That’s the part that doesn’t make it into the slide deck.
What’s being promised
Meanwhile, the asks keep escalating. AI-driven X. AI-powered Y. Roadmaps assuming AI will handle 30% of feature work by next quarter, before anyone has actually validated that claim against a real codebase with real users and real edge cases.
This isn’t a complaint about ambition. It’s an observation about a particular kind of risk: when the people setting timelines have only seen AI in demos, and the people doing the work have seen it both succeed and fail in production, there’s a translation problem.
The honest take
Both rooms are partially right.
Engineering is right that the gains are uneven, the tools have rough edges, and most of the breathless claims won’t survive contact with production. We’ve seen this movie before — every wave of new tooling promises 10x and delivers 1.3x for the team willing to adapt and 0x for the team that doesn’t.
Business is right that something real is happening. AI isn’t another framework trend that’ll fade. The capability shift is genuine, even if the timeline isn’t.
The teams that win the next two years aren’t the ones with the loudest AI strategy. They’re the ones that treat AI like every other tool — with healthy skepticism, real measurement, and patience for the gap between “this works in a demo” and “this works at 2 AM during an incident.”
Promise less. Ship more. Measure honestly.