# The Work That Runs Without You

*Published 2026-10-15* | Author: brad-webb

> **tl;dr** By hand, a trial that went quiet on Tuesday heard from me nine days later, or not. A rule drafts the check-in at ten past ten. This part sorts the small customer work onto three rungs by what one wrong run would cost, and writes the rule in four fields. Part 3 of 4 of The Growth Team of One.

*This is part 3 of 4 of The Growth Team of One. Part 1, [The Job Nobody Wrote Down](https://www.civic.com/field-notes/the-job-nobody-wrote-down), wrote the job down. Part 2, [The Work You Stop Doing](https://www.civic.com/field-notes/the-work-you-stop-doing), decided what to stop and circled what never stops. This part hands the circled lines to a rule. Part 4, The Stack You Hand Over, is next Thursday.*

Part 2 ended on the short list of work that never stops: the things a customer would notice by lunch if they slipped. Five lines landed on mine. The follow-up when a trial starts. The reply to the inbound form. The note when a known account comes back to pricing. Finding out that an integration broke. The case study sales asked for. Each one is small, triggered, and unasked for, and each one is the first thing a customer feels when it slips.

Here's the part I'd rather skip. I kept doing all five by hand for about a month after I knew a rule should be doing them. I'd written the sentence myself, more than once: hand the small customer-triggered lines to a rule. Then I went back to sending the trial note from my own inbox at whatever hour I noticed the trial, which was sometimes Thursday.

The reason, when I finally said it out loud, was that I'd been asking the wrong question. I was asking whether I trusted it. That question has no answer, because trust is a feeling about the future, and you can argue with yourself about the future for as long as you like. The question with an answer is smaller. If this ran wrong once, who would notice, and what would it cost? That's a question about one line and one bad morning, and you can answer it in a breath.

## A rule has no calendar

I'll say what a rule is first, in plain words, because the word arrives with baggage.

A rule you set once: when this happens, do that. Bryn, the GTM Agent we make at Civic, calls that a Play, and I'll use the plain word for the rest of this piece. The "when" is a signal: something a buyer just did, like coming back to your pricing page, or starting a trial and going quiet. The "do that" is the small response you were doing by hand. A rule has four parts: when this happens, do this, from this owner, and write down that it ran.

Here's what makes it different from you, and it isn't that it's faster. A rule has no calendar. It has no slot to protect on Thursday, because it doesn't need a slot. It has no bad week, because it doesn't have weeks. It doesn't notice the trial at four o'clock because that's when you had a gap. It notices at the minute the trial goes quiet, and the minute is the whole point. These small customer lines are worth the most at the moment they fire and lose value by the hour, and the one thing a person can't do is be free at an arbitrary minute. A rule is free at every minute. That's the entire argument for it.

Now the limit, stated as plainly. A rule can't judge. It can't tell that this trial is the founder's old colleague and wants a call, not a note. It can't tell that this account is halfway through a renewal argument and the pricing visit means something else. It does what you wrote, every time, including the time you wouldn't have. That isn't a flaw to fix. It's the reason the next question matters.

## Three rungs, by what goes wrong

Take each line on that never-stop list, the work a customer would notice by lunch, and ask: if this ran wrong once, who would notice, and what would it cost? The answer lands on one of three rungs.

**It runs on its own.** Wrong once costs nothing a customer sees. The rule tags the account, writes the note in the deal channel (the Slack channel where the account lives), opens the task. If it misfires, you delete a task. Noticing that a known account came back to pricing is this kind of line: the noticing is internal. So is finding out that an integration broke, because a false alarm in your own channel costs you a glance.

**It runs after my yes.** Wrong once is an awkward message to a real person. The check-in to a trial that went quiet. The reply to the form. The case study that goes to a buyer. The cost of a miss is an email you wish you'd read first, so you read it first. Bryn drafts the step and waits for your yes. That's Approve mode, and it's where almost everything on that list starts and where a good deal of it should stay.

**It stays mine.** Wrong once is a lost deal or a price. The call about whether to chase the deal that went quiet. The number you quote when it isn't on the list. The conversation with the customer who's about to leave. Nothing on my never-stop list landed here, and that's not an accident. Back in part 1, when I wrote the whole job down, the judgment lines were tagged "only you", and they never reached this list, because they aren't small and they aren't triggered. If one of your five lands here, look again. Most lines that seem to belong on this rung are two lines stapled together: the part that notices, which can run, and the part that answers, which can't. Chris, our CEO, made a related point yesterday about buyers whose research is done before a person ever shows up to verify it. The verifying is the judgment line, and it's yours.

Notice what sets the rung. Not how much I trust software. Not how clever the rule is. What a single wrong run would cost, and who would pay it. The same line can sit on a different rung at a different company, and that's correct.

WHERE EACH LINE CAN REACH (interactive: the five lines a customer would notice by lunch; for each, answer what one wrong run would cost, and the line moves to the rung that answer points at)
RUNS ON ITS OWN (if it ran wrong once: nobody notices):
- Notice when a known account comes back to pricing (Sales assist)
- Find out when an integration breaks (Tools)
RUNS AFTER MY YES (if it ran wrong once: a customer gets an awkward message):
- Follow up when a trial starts (Lifecycle)
- Triage form fills and the inbound inbox (Sales assist)
- Send the case study sales asked for (Sales assist)
STAYS MINE (if it ran wrong once: a customer loses something real)
The default placement is illustrative, one possible read. Yours will differ.
Figure 1. The five lines a customer would notice by lunch. For each, answer what one wrong run would cost, and the line lands on its rung. Illustrative defaults; your read will differ.

## Writing the rule

A rule is four fields, and if you can't fill all four it isn't a rule yet.

**When this happens.** The signal, specifically. Not "a trial needs attention" but "a trial has had no activity for three days." If you can't name the moment, the rule will fire on the wrong one.

**Do this.** One step. Send the check-in. Post to the deal channel. Create the task. If you want three steps, write three rules, so that each can sit on its own rung.

**From this owner.** Whose name is on it. The check-in comes from the account owner, not from "the team," because the customer replies to a person.

**Write it down.** Always. Every run leaves a line: what fired, what it did, from whom, and whether a person said yes first. The record isn't paperwork. It's the only way you'll know, in a month, whether the rule is doing what you meant.

ONE RULE, FOUR FIELDS (interactive: pick a line, then fill the fields; the card reads the rule back as a sentence and as the record line it would leave)
Lines: Follow up when a trial goes quiet; Notice when a known account comes back to pricing; Spot a second person from the same company.
When this happens: trial goes quiet 3 days; second person from the same company; pricing page, third visit this week.
Do this: send the check-in; post to the deal channel; create a task.
From this owner: me; the account owner; the team inbox.
Write it down: always (fixed: every run leaves a line).
Worked example. Line: Follow up when a trial goes quiet. When a trial goes quiet for 3 days, send the check-in from the account owner, and write down that it ran.
Record line (illustrative): run ⬩ trial goes quiet 3 days ⬩ send the check-in ⬩ the account owner ⬩ logged
Figure 2. One rule, four fields. Pick a line, fill the fields, and read the rule back as a sentence and as the record line it would leave. Illustrative.

The fourth field is the one people skip, and it's the one that makes the other three safe to hand over. A rule you can't read back is a rule you can't promote, because there's nothing to read.

## The calendar versus the rule

Here's what the month by hand looked like. I had a slot on Thursday afternoon called "trial follow-ups." It was protected, in the sense that it was in my calendar in a color. Most weeks something landed on it. The slot would slide to Friday, and on a bad week to the following Thursday, and a trial that went quiet on Tuesday would hear from me nine days later, or not.

The rule has no Thursday. The trial goes quiet on Tuesday at ten past ten, the rule drafts the check-in at ten past ten, and the only open question is whether I say yes before lunch. Same line. Same words in the note. The difference is that one of them waited for me to be free and one of them didn't. I wrote about the by-hand version of this line in [Follow-Up Has No Owner](https://www.civic.com/field-notes/follow-up-has-no-owner); the rule is what an owner looks like when the owner isn't a person.

ONE WEEK, TWO ROWS (interactive: row A is the calendar in one busy week, preset; row B is the rule; fire the signal on any day, at any hour, and the rule answers there)
A. The calendar (slot: Thu 4pm, trial follow-ups): Mon free; Tue free; Wed free; Thu busy (the 4pm slot slid to Friday); Fri 4pm trial follow-ups (slid). On a worse week it is the following Thursday.
B. The rule (when a trial goes quiet: draft the check-in): the trial went quiet on Tuesday at 10:14; the rule drafted the check-in at 10:14, the same day and hour the signal fired.
Illustrative clock and an illustrative week.
Figure 3. One week, two rows. The calendar row is preset to a busy week: Thursday was taken, so the protected slot slid to Friday. Fire the signal on any day at any hour and watch the rule answer there. Illustrative clock.

## Promotion, and the log that earns it

Nothing starts on the top rung. Every rule I've written started where it waits for my yes, including the ones whose wrong run would cost nothing, because the first few runs aren't about the cost. They're about finding out whether I wrote the "when" tightly enough. I read every draft. Most were right. After a few weeks the yeses got boring, which is the signal you are waiting for.

The wrong drafts were the useful ones. Each showed a rule written too loose. The check-in that drafted for a trial whose company already had an open deal with us: that one taught me to add a line to the rule, not to stop trusting it. The rule got tighter and the next run was right. Tuesday's piece made the same point from the vendor's side, [on by default is not approved by default](https://www.civic.com/field-notes/on-by-default-is-not-approved-by-default). Approval is something you give a rule after you've watched it, not something it arrives with.

So promotion is simple, and it isn't a feeling. Read the runs. When you've said yes to a stretch of them without editing one, move the line up a rung. If the next rung's wrong run would cost a customer something, the line stays where it is, and that's the rung it lives on. The log is the trust. Not the demo, not the vendor's promise, not my gut. Five runs I can read.

ONE LINE, FIVE RUNS (interactive: the line starts where it waits for your yes; reveal the runs one at a time, read each, and decide when it moves up)
Line: Follow up when a trial goes quiet. Sits on: runs after my yes.
- run 01 ⬩ Mon 09:41 ⬩ fired: trial quiet 3 days (account A) ⬩ drafted: check-in from the account owner ⬩ decision: yes
- run 02 ⬩ Tue 10:14 ⬩ fired: trial quiet 3 days (account B) ⬩ drafted: check-in from the account owner ⬩ decision: yes
- run 03 ⬩ Wed 15:52 ⬩ fired: trial quiet 3 days (account C, open deal) ⬩ drafted: check-in from the account owner ⬩ decision: edited. Note: this showed me the rule was too loose: added the suppression for open deals
- run 04 ⬩ Thu 08:07 ⬩ fired: trial quiet 3 days (account D) ⬩ drafted: check-in from the account owner ⬩ decision: yes
- run 05 ⬩ Fri 13:30 ⬩ fired: trial quiet 3 days (account E) ⬩ drafted: check-in from the account owner ⬩ decision: yes
Illustrative runs and illustrative clock times. The edited one is the useful one.
Figure 4. One line, five runs. Read each run and decide when the line moves up. Illustrative runs; the edited one is the useful one.

## What the record already shows

Bryn is the product we make for this, and we ran it on ourselves first, and we still do. [We Are Customer Zero](https://www.civic.com/field-notes/we-are-customer-zero) tells that story. Everything above is how I came to run it: every rule I wrote started in [Approve mode](https://www.civic.com/field-notes/approve-mode-is-not-a-speed-bump), where Bryn drafts the step and waits for my yes; every run is logged, what fired, what it did, from whom, and who said yes; and any rule stops with one click, which is most of why I was willing to let go of the wheel at all. If you want the parts of a rule taken apart one at a time, [Anatomy of a Play](https://www.civic.com/field-notes/anatomy-of-a-play) does it in a page. Bryn is not another dashboard to watch. It is the governed execution layer that runs Plays through your stack. In plain terms: it runs the rules you wrote, through the tools you already use, and keeps the record of every run. The rungs are yours to set. The log is how you'll know you set them right.

## The gift

Take your never-stop list from last week, or the five above.

For each line, answer one question: if this ran wrong once, who would notice, and what would it cost? Nobody and nothing: it runs on its own. A customer, and an awkward email: it runs after your yes. A customer, and a lost deal or a price: it stays yours.

For every line on the first two rungs, write the rule in four fields. When this happens. Do this. From this owner. Write it down.

Start every rule where it waits for your yes. Read its runs. Move it up one rung only after you've read them and stopped editing them. Never move it past the rung its wrong run can afford.

THREE RUNGS, BY WHAT GOES WRONG (static, illustrative)
The rung is set by what one wrong run would cost, and who would pay it.
RUNS ON ITS OWN (if it ran wrong once: nobody notices):
- Notice when a known account comes back to pricing
- Find out when an integration breaks
RUNS AFTER MY YES (if it ran wrong once: a customer gets an awkward message; every rule starts here):
- Follow up when a trial starts
- Triage form fills and the inbound inbox
- Send the case study sales asked for
STAYS MINE (if it ran wrong once: a customer loses something real):
- Quote a price not on the list
- Decide whether to chase a quiet deal
- Talk to the customer about to leave
Illustrative: your lines will differ, the test will not.
Figure 5. The three rungs, with the test under each. Every rule starts on the middle one. Illustrative: your lines will differ, the test will not.

## Where this goes

Part 4, The Stack You Hand Over, next Thursday: the day the whole chart gets handed over with a second name on it, and the record becomes the onboarding doc.

*Part 3 of 4 of The Growth Team of One.*

**Stop watching signals. Start running them.**

Source: https://www.civic.com/field-notes/the-work-that-runs-without-you