Why slowing down first makes engineering faster

Engineering teams are producing more code than ever with AI, and too often the business doesn’t feel it, because the extra work gets lost before it reaches customers. Getting more of that work to your customers starts with slowing down, which surprises some CEOs.

What slowing down means

When I say slow down to speed up, I mean putting the foundations in place before you push for more: each engineer trusted to make the calls in their own area, clear guidelines and responsibilities, ownership in the hands of the right leaders and the people doing the work, and code review that is set up and expected. AI can’t put any of that in place for you, and without it more code mostly means more for someone to check.

Going faster, or only believing you are

The test I use for real speed is simple: on the teams I’ve led, nothing counts as done until it works on real data with real customers. AI makes it easy for a team to trade its own judgment for the feeling of speed. Someone has AI write up every meeting so everybody knows what to do, and it becomes one more document that disappears off into forever.

Slowing down also leaves room for junior engineers to learn the system while AI does more of the typing, and the engineers who know your systems are the ones who can bring them along.

Why it matters more now

The research keeps finding the same gap between what engineering produces and what reaches customers. In LeadDev’s 2026 survey, 52% said code review and validation is now the biggest bottleneck in delivery, against 7% who said writing code is. Faros, which sells engineering analytics, studied 22,000 developers for its September 2026 report and found the average change about 72% larger than before, with 76% more changes merged without any review. DORA’s 2026 report on AI returns puts it plainly: without the right organizational foundation, “AI creates localized pockets of productivity that are often lost in downstream chaos.”

Where the first weeks go

When my team and I start with a CEO, the first weeks go to finding where the gains from AI leak out, choosing one focus with a named owner, and starting on the handoff costing you the most, where work passes from one person or team to the next, with your CTO in charge of the team throughout. That’s the slow part, and by the end of it you can see whether the first signals are showing up. If your team is shipping more and the dates still slip, bring the last date that slipped to a thirty-minute call, and we’ll trace where the time went.

Read next

What to bring when your board asks if AI is paying off

All writing