I’ve been spending a lot of time learning how to use AI in my job and helping other managers do the same. There’s one failure mode I’ve witnessed recently that traps quite a few folks and the root cause is that they’re asking the wrong question: “Did AI help me do this faster?” Or similarly phrased as a complaint “This takes me even longer when I use AI.”
This velocity trap is easy to fall into, especially for those of us who manage software engineers. You might observe your team use AI coding assistants and, from a distance, conclude the only (or most important) benefit is speed. That’s an uninteresting takeaway from AI-assisted coding and it’s a bad lens to carry into your own work.
Managers rarely get stellar feedback loops, so you often can’t tell if you’re doing the job well. When you can’t measure quality, speed is the only thing that’s left and so velocity becomes the default frame. That means it’s on us to define high standards and track progress towards them.
For managers, the big lift is “How can I do my job better?” I’m writing my tenth performance review of the cycle. If I’m focused on velocity I’ll ask how fast the drafts came out. If I’m focused on quality I’ll ask whether the feedback was specific, evidence-backed and actionable. Did the engineer walk out knowing what to do different or what to do more of?
Successfully leading empowered teams is based on a foundation of coaching and trust. An efficient manager can still be an ineffective coach. Focus your AI use on being a better manager and coach for your engineers, not how much less time you spend prepping for 1on1s or drafting development check-ins.
Next: the first instinct you have to unlearn to make that work.