When I would rather ask an agent than hire a person

There are moments now where I would rather talk to Claude Code or Codex than hire a person.

That sounds harsher than I mean it.

I am not saying humans are less valuable. I am saying the baseline for what justifies a hire has changed.

Let me make it concrete with a typical feature workflow.

I explain an idea. Notes get turned into a task. The team builds a first version. Then the real iteration starts. I review, suggest design changes, ask for copy tweaks, notice navigation issues, push on task flow, request better UX. The team goes back, raises a pull request, I review again, and we deploy.

That loop is familiar. And AI is changing large parts of it.

Now I can turn the same notes into a spec, ask an agent to build a first pass, comment and redirect within the same session, and get to something reviewable much faster. The loop is not gone. My responsibility in it is actually more visible. But what I need from a human hire has shifted.

The question I ask now is not “who can do this work?” It is “what would still sit with me if I had a good agent for the first pass?”

The honest answer is: the parts that require judgment, taste, and accountability.

• Can they translate an ambiguous brief into a clear problem statement with real acceptance criteria?
• Can they see that a feature is functionally correct but weak on task flow or information hierarchy?
• Can they identify when the issue is not a component tweak but a navigation model problem?
• Can they review a pull request with product context: edge cases, regression risk, whether the implementation matches the intent?

Most importantly: can they reduce the number of times a decision comes back to me?

That is the real bar. Not output speed. Not AI proficiency in the abstract. The ability to carry more of the loop before it returns to the founder.

The uncomfortable question for any hiring manager right now is: would I still hire this person if I already had a capable agent for the first pass?

If the answer is no, the role is probably too execution-heavy for what it needs to justify.

If the answer is yes, name why. Exactly why. Because that answer. Specific, honest. Is what the role should be built around.

In the AI era, “can do the task” is too low a bar. The better hire is someone who can make the task worth doing, make the output worth trusting, and take enough ownership that the founder can focus on the next layer of the business.

Agents do not remove accountability. They make it more obvious where accountability was missing in the first place.