The more I work with AI as an engineering manager, the more I find one skill has become dramatically more valuable: the deliberate pause.
I know that might sound backwards. We're surrounded by tools that let us build faster, write faster, analyze faster, and ship faster. With each new advancement, the expectation seems to be that we should be producing more in less time. It feels like the people moving the quickest must be winning.
But AI hasn't eliminated the biggest bottleneck. It's just made it easier to reach it.
Us.
The hardest part of most work isn't implementation anymore. It's deciding what deserves to be implemented in the first place.
AI can help you build a feature in half the time, but it can't reliably tell you whether it's the right feature. It can generate a beautiful report in minutes, but it can't tell you whether anyone should have asked for that report.
We've made tremendous progress in how quickly we can produce things. I'm not convinced we've made the same progress in deciding what we should produce.
When Generation Becomes a Distraction
One of the traps I've started noticing is how easily we confuse generation with progress.
One of AI's greatest strengths is making exploration incredibly cheap. That's a good thing! If my team can explore 50 potential solutions in the time it used to take us to explore five, I'll happily take that trade.
But there's a cost we don't always account for.
Every solution we generate is another solution competing for our attention. Someone still needs to understand what was produced, evaluate whether it solves the problem, and decide what to do with it.
If 45 of those solutions were never worth considering in the first place, we've also created 45 distractions. We've made the search harder.
This becomes especially important when working on a team. Generating something doesn't mean it's ready to be used. Someone else may need to review it, maintain it, or build on top of it. The time we save during implementation can easily become time someone else spends figuring out what we've created.
That's not an argument against using AI to explore more possibilities. Exploration is valuable, and being able to discard an unsuccessful approach quickly is a tremendous advantage.
The problem comes when we treat the number of things we've produced as evidence of how much progress we've made.
Bringing Back the Deliberate Pause
This is why I keep coming back to David Marquet's idea of the deliberate action and it's use of a mandatory pause.
There's an important distinction between pausing and slowing down. We tend to associate slow with failure. If something isn't moving quickly, we assume there's a problem. But a pause isn't an absence of progress. It's an intentional moment to make sure we're heading in the right direction.
In engineering, we're familiar with the cost of making the wrong decision. A feature that solves the wrong problem doesn't become valuable just because we delivered it ahead of schedule. An implementation that creates unnecessary complexity doesn't become easier to maintain because it took less time to write.
The deliberate pause creates an opportunity to ask a few questions before committing our energy:
- What problem are we actually trying to solve? It's easy to jump into a solution before we've fully understood the problem. AI makes that temptation even stronger because we can start generating something almost immediately.
- Is this the highest-leverage thing we could be working on? Just because something is easy to produce doesn't mean it's worth producing. We still have to decide where our attention will make the biggest difference.
- What would success actually look like? Without a clear definition of success, it's difficult to distinguish a useful result from something that merely looks finished.
These aren't complicated questions. They're questions we already know we should be asking. The challenge is remembering to ask them when the tools make it so easy to skip ahead.
Five minutes of thinking can save hours of implementation.
Consistency Builds Trust
Ironically, AI makes that pause even more valuable.
When implementation was expensive, thinking was naturally built into the process. You had time while you were writing code, creating documents, or building presentations to reconsider what you were doing. That didn't guarantee good decisions, but the work itself created opportunities to reflect.
Now implementation is almost free.
Good judgment isn't.
As an engineering manager, I'm less interested in how many things someone can produce than in whether I can consistently trust the results of their work.
That doesn't mean speed isn't important. Delivering quickly matters, and someone who takes forever to make a decision isn't necessarily exercising better judgment. There's a balance between moving quickly and making sure the work is worth doing.
But consistency is what builds trust over time.
When someone consistently delivers useful work, the people around them learn they can depend on it. They spend less time second-guessing decisions, questioning whether something is ready, or worrying about what they'll discover later.
That trust makes teams more effective. It also makes you the person others want to work with when something really important needs to get done.
I suspect the engineers who stand out over the next few years won't necessarily be the ones producing the most output. They'll be the ones who can consistently turn these increasingly powerful tools into useful outcomes.
Finding the Needle
AI has made it incredibly easy to create haystacks.
We can generate more code, more documents, more solutions, and more possibilities than ever before. And there's genuine value in that. Being able to explore more options without investing enormous amounts of time is an opportunity we shouldn't ignore.
But quantity alone doesn't guarantee we're getting closer to what matters.
If our goal is finding needles, creating more hay isn't automatically progress. At some point, we have to stop adding to the pile and consider whether we're looking in the right place.
That's where the deliberate pause earns its value. Not by replacing speed, but by making sure that speed is taking us somewhere worth going.
The goal was never the haystack.
It was always the needle.