Essays/Leadership

Delegation starts with clarity

What cannot be explained exactly cannot be delivered. Delegation is collaboration that starts with clarity, and doing it badly costs decisions as well as money.

If I cannot explain exactly what I want, nobody can deliver it. Delegation is not a way to get rid of work I dislike. It is collaboration that begins with my own clarity, and without that clarity it costs time, money and decisions.

Why do the work yourself first?

Because it buys the knowledge you need to delegate well. Founders often look at a website, a product or a design and believe they could make it at least as well themselves. They probably could. The better question is whether they want to, because time is the one resource nobody gives back.

Early on, when money is tight, I think doing the work yourself is usually the better choice. It saves money, and it earns judgment. You cannot judge a designer’s work if you have never wrestled with design, or tell a good marketing hire from a confident one if you have never written a campaign. Doing the work badly at the start is how the judgment gets earned. Founders who delegate everything from the first day spend the rest of the company’s life unable to tell whether the people they hired are any good.

So delegation should not begin without knowledge, and the knowledge in question is my own. I have made this mistake. Wanting to move faster, I put money into partnerships I hoped would speed up my projects. They did not, because I had not understood what I wanted from those projects. When I cannot explain what I want, people do the work their way. Whether the result matches what I had in mind then depends on their personality, their experience and their own intentions for the project.

A good partner works from a simple description

I think of delegation as collaboration. The other person should be able to bring their own contribution, by paths that differ from mine, and still arrive at the finished project.

A good partnership is one in which the partner can produce what I have in mind from a simple description.

Two counterexamples. If I work with a recruitment agency, my condition is that I see the last, best candidate. I should not end up refining CVs and talking to people until I reach the result myself. If I work with an events agency, a bad process is one where they send me everything, from arrangements to chair types, and I sit and choose. A good process is one where I describe what needs to happen and they make it happen, with a few small questions along the way.

I broke my own principle recently. An agency was to do the branding for one of our companies, and I sat through seven hours of calls with them. They asked question after question, very technical, the kind learned in school: a lot of system and structure, and none of the refinement a brand needs. After seven hours they produced a disastrous mock-up. I ended the collaboration before it went further, paid part of what we owed, and walked away. I think it is right to pay what you owe, sometimes a little more than strictly necessary, because it keeps the door open.

Complaining to myself about a partner changes nothing. What works is telling them on the first day, kindly and in full sentences, what I expect and what result I am after, in detail but without constraining how they get there. When it is clear they cannot meet that expectation, it is time to end the collaboration and find someone else.

Things cost decisions, not just money

There is a cost in delegation that is easy to miss. When a business owes a large sum and cannot pay it, the money is not the real problem. The problem is the decisions it is forced to make in order to pay: the client that should have been declined, the project that should have been closed, the terms that should have been rejected. The money is recoverable. The chain of compromised decisions is what sets a business back.

This is why long contracts are dangerous early. I have negotiated long contracts, plans changed, and a day came when I no longer needed what I had signed for and still had to pay. So early on I would keep collaborations temporary, with the shortest contract life and the shortest notice period available.

The same rule with AI agents

Much of my delegation now goes to AI agents, and the rule has not changed. I tell a coordinating agent the outcome I want and leave the implementation to it: you decide the approach, but keep it scalable. The how is the layer that should not need me. The what always does, and it is the one decision I cannot hand over.

An agent’s default is to act, not to ask, which is the events-agency rule again. The exception is scope. An agent that starts building something I only asked to discuss gets stopped, in writing, every time. How I state what I want is the subject of Voice-first specs. What belongs here is the check afterwards. A standing duty handed to an agent is followed by a request for something I can verify, because a report that the work is done is not the work, as Agent self-reports are not evidence explains.

Delegation done well buys time. Done badly, it spends the same time twice.

Back to the essays

Field Notes

A letter from Andrei Hasna on building with AI agents

Subscribe by email

By clicking “Subscribe”, I confirm that I want to get email delivered to my address and agree to Andrei Hasna’s Privacy Policy and Terms.

In each letter

  • New essays on AI agents and the systems behind them
  • What Hasna is building, and why
  • Costs, routing and results from real work
  • What did not work
  • News of the book and the podcast