How to Build Agency When You're Smart Enough to Wait Forever

There is a particular kind of delay that looks good from the outside.

You spend three days thinking about a design decision that one message could clarify. You keep polishing a draft nobody has seen. You wait until you are nearly blocked before asking for help, because asking earlier would mean admitting what you do not know.

Sometimes that is care. Sometimes it is fear in a very respectable outfit.

By agency, I mean the habit of shortening the gap between noticing a problem and taking a useful step. You ask, share, or test something that lets you learn what private thought cannot settle.

The before-lunch rule

For the next two weeks, try one thing every workday that creates an opportunity for feedback before lunch.

Send the rough draft. Ask the person who owns the requirement. Test the assumption that has held up your design. Offer to write the first half-page of a plan.

Choose something that can change your understanding. Sending a status message for the sake of sending one does not help. Neither does polishing the same private document for another hour.

At the end of the day, write three lines:

What did I notice?

What did I do?

What did I learn, or what response am I waiting for?

The third line matters. Activity alone is a poor measure of progress. You want to learn whether an earlier action resolved uncertainty, exposed a missing fact, or gave someone else enough time to help.

Make the next question concrete

Suppose you are choosing between two designs for an import feature. You could spend the afternoon comparing them. But the choice may depend on whether users need to correct individual rows or resubmit the entire file.

That is a question for the person who understands the workflow. A short note can make it answerable:

“I have two possible designs. The choice depends on how users correct a failed import. Do they need to fix individual rows? I have sketched both flows so we can check.”

This is an illustrative example, but the distinction is practical: identify the missing fact, then choose an action that can supply it. “I need more context” is hard to finish. A named question has an answer, an owner, or an explicit uncertainty you can discuss.

Three moves worth practising

Ask while the answer can still change the work. Explain what you understand and which fact you need. You do not need to wait until all progress stops.

Show a draft with a question attached. Roughness helps when it invites a useful correction. “Does this cover the failure case?” gives the reader a better task than “Any thoughts?”

Take the first part of ambiguous work. Offer to frame the options, test an assumption, or identify who needs to decide. You can take responsibility for that step without promising to solve the whole problem.

Each move involves a small amount of exposure. A question may reveal a gap in your knowledge. A draft may prove your first idea wrong. That can be uncomfortable, especially when people rely on you to produce considered work. The benefit is discovering the problem while changing direction is still cheap.

Give a useful pause an exit condition

Some actions affect people who did not choose the risk. A production change, public promise, or staffing decision deserves more care than a private sketch.

Before acting, ask:

1. Can I reverse this step?
2. Who bears the cost if I am wrong?
3. Is there a smaller step that would answer the question first?

Sometimes the right action is to wait for a decision. Make the reason and the next event clear:

“I am waiting for the contract owner to confirm whether clients rely on this behaviour. If I have not heard by Thursday, I will ask for a decision on the assumption.”

Compare that with “I need to think about it a bit more,” renewed each afternoon. The first pause protects a known dependency. The second may have no way to end.

The before-lunch rule is an exercise, not a deadline for other people. Share work early enough that they can respond on a reasonable schedule. If you already know what to do and need uninterrupted time to do it, protect that time.

At the end of two weeks, look for one useful change: did an important conversation happen earlier? Keep the action that helped make it happen.

Taking the first step early and checking a consequential change carefully can fit together. The Velocity Paradox looks at that balance in AI-assisted engineering.

Revised 5 September 2026: moved the exercise earlier, combined repeated advice, and added a clearer way to assess whether an action helped.