Station 05 · The Choosing Table

Pick the right first agent.

Most projects fail because the first agent was the wrong agent. We'll use a small matrix, four properties, and five questions to narrow your shortlist to one.

The picture

Effort vs value. Start in the top-right.

Plot your candidates. Build the one with the highest value and the lowest effort.

🤖

Pip thinks…

The most common mistake here is the high-value, high-effort agent. The "obviously useful" one that's a four-month build. Save those for agent #3.

The four properties

What makes a good first one.

All four should be true. If three are true and one is missing, you've found which gap to fix first.

Pip beside a small floating loop-arrow icon.

1. Repetitive

It happens at least daily, and the answer is roughly similar each time.

Pip holding a neat stack of clean labelled papers.

2. Clean data

The inputs already live somewhere accessible. Don't fix data and build agents in one project.

Pip catching a falling envelope on a small pillow.

3. Safe to fail

A wrong output causes embarrassment, not financial or legal harm.

Pip beside a small stopwatch and chart card.

4. Measurable

You can count the hours per week it saves, in numbers, by end of month one.

The five questions

One final filter.

Run your shortlist through these, in order. Stop at the first "no" and pick a different candidate.

Pip beside a small calendar with 3+ days circled.

1. Does this happen at least 3 times a week?

If no → reconsider. The win compounds with frequency. Once-a-month tasks rarely justify the build.

Pip pointing at three small labelled boxes.

2. Can I describe the data sources in one sentence each?

If no → the data isn't ready. Fix the plumbing first as its own project, then come back.

Pip resetting a stopwatch beside a fallen stack of envelopes.

3. Is the worst case recoverable in under an hour?

If no → skip for the first agent. Build something where embarrassment is the worst case.

Pip sketching the shape of a message on a paper.

4. Can I write the exact format of the output?

If no → you're not ready. If you can't describe the output, you don't know the task well enough to delegate it.

Pip placing a mint checkmark on a week-1 calendar.

5. Will I know in week one whether it's working?

If no → measure first, build later. Most agent failures are actually measurement failures.

Five yess. That's your first agent.

Try this, before you scroll on

Run your shortlist through all five questions.

Take the two candidates you picked at Station 04. Walk each one through all five questions. Where did the first "no" come for each? The one with the fewest gaps wins.

You finished Station 05

Wise Picker

Fifth sticker. Four more to collect.