Chat or hand-off: telling the two kinds of work apart

1Two different shapes of help

By now you've done the first kind: you type something, you get something back, you judge it, you type again. You are holding the loop. Every step passes through you, and nothing happens that you didn't just ask for.

There is a second kind, where you describe a finished outcome and the tool goes away and takes multiple steps on its own — reading files, making intermediate decisions, checking its own work — and comes back when it's done or stuck. The tool is holding the loop.

Explain-on-first-use: the second kind is usually called agentic. It means the same as it does in "estate agent" — something acting on your behalf, making small decisions you didn't individually approve, toward an outcome you specified. That's the whole idea. The vocabulary is more intimidating than the concept.

Neither is better. They suit different jobs, and picking wrong is expensive in both directions.

This section sits here, before the two tool sections that follow, because the question it answers — am I holding the loop, or handing it over? — is the one you'll be asking of every tool in Sections 3 and 4.

2The test

Before you start, ask two questions in this order.

1. Can I describe the finished thing without describing the steps?

"Ten hooks for this product" — yes. One step, one output. "Go through my last thirty posts, find every one that outperformed the median, and tell me what the openings have in common" — also yes, and it's five or six steps you'd rather not do by hand. That's a hand-off. "Make this funnier" — no. You can't specify the outcome, so you can't delegate it. Stay in the conversation.

2. Will I be able to tell if the result is wrong?

This is the one that actually decides it, and it's the one people skip.

If the tool renames two hundred files and gets nine wrong, you'll spot it — the work is checkable at a glance. Hand it off.

If the tool reads thirty competitor posts and reports a pattern, you have almost no way to check whether that pattern is real or whether it built you a confident-sounding summary of nothing. That doesn't mean don't do it. It means treat the output as a hypothesis you go and verify against three actual posts, not as a finding.

Delegate the work you can check. Stay in the room for the work you can't.

3Where each one earns its place in UGC work

Stay in the chat for:

• Hooks, lines, rewrites — anything where you're judging each output. Section 1 is entirely this kind of work.
• Anything where the criterion is taste. Funnier, warmer, less like an ad.
• Fact-checking a specific claim before it ships, where you want to see the reasoning rather than a verdict.
• Thinking out loud about an angle. The value is the back-and-forth, not the output.

Hand it off for:

• Batch work against a fixed standard. Twenty scripts checked against the off-limits list from Form BV-01. Mechanical, checkable, tedious — the ideal profile.
• Anything involving many files. Renaming, sorting, pulling every clip that mentions a product, assembling a contact sheet of the week's footage.
• Repeated pipelines you've already done by hand and can describe. This is the exact boundary where a hand-off becomes worth turning into a Skill, which is Module 5.
• Research sweeps you intend to verify. Competitor patterns, brand tone audits, gathering the last six months of a brand's captions in one place.

Do neither for:

• The final call on which take ships.
• Anything that goes out under a client's name without you reading it. The temptation at volume is real, and it is precisely how a brand ends up saying something it's contractually barred from saying.

4The failure mode nobody warns you about

Chat and hand-off fail differently, and the difference matters more than the capability difference.

A chat fails early and visibly. You ask for a hook, you get a bad hook, you have lost eleven seconds and you know immediately.

A hand-off fails late and confidently. It works for twenty minutes, makes a wrong assumption in step two, builds nine further steps on top of it, and hands you a complete, well-organised, internally consistent result that is wrong from the second step onward. Nothing about the output looks distressed. That's the danger — the presentation quality is unrelated to the correctness.

Three habits, in order of how much they save you:

1. Run it once on five items before you run it on two hundred. Almost every expensive hand-off failure would have been caught on a sample that took two minutes.
2. Ask it to state its assumptions before it starts. A single line — tell me how you're interpreting this before you do it — catches the step-two error while it's still free.
3. Check the first and last item of any batch by hand. Not a random sample — specifically the first and last, because drift tends to show at the ends.

5What this costs

Hand-offs consume far more work per request than a chat does, and they take longer. That's the trade: you're buying back attention, not time, and not always time.

For a beginner the honest advice is to stay in the chat longer than feels efficient. You cannot supervise a process you have never performed. Do the twenty scripts by hand once — badly, slowly — and you will know what a wrong one looks like. That knowledge is the entire prerequisite for delegating it, and there is no shortcut to it that survives contact with a client.

6Exercise 2.2 — Sort your own week

Don't run anything for this one. Just sort.

List every task you did on your last piece of content, from first message to delivered file. Ten to twenty lines is normal.

Against each, mark:

• C — needs a conversation, I'm judging each output
• H — could be handed off, and I could check the result
• M — mine, and staying mine

Then look at the H column and ask the honest question: have you ever done that task by hand? Anything in H that you've never done manually gets moved back to C until you have. That's not caution for its own sake — it's that you can't spot a wrong answer to a question you've never worked through.

The M column, incidentally, is your actual job description. Most people find it's shorter than they expected and more valuable than they thought.