The morning sweep: what has gone quiet
Nobody notices a deal going quiet; they notice a quarter going quiet. One workflow, every morning at half past nine, turns silence into a list of names.
Every workflow so far has reacted to something — a lead arriving, a stage changing, a form coming back. This one reacts to nothing happening, which is harder, because nothing happening does not raise an event.
It runs at 09:30, 6 days a week, and asks one question: what has gone quiet? An enquiry nobody has touched is the most expensive thing in a pipeline. It costs nothing to keep, it flatters every report it appears in, and it is dead.
What it does, in one picture
7.2 · Every morning, chase what has gone quiet
A clock, a query, a loop that raises one task per quiet enquiry, and — after the loop, not inside it — a second look for the ones that are properly serious.
Scroll for all 7 steps →
How it measures silence
This is the part worth stealing. The query does not ask when the record was last updated — it asks when somebody last wrote a note on it, falling back to when the enquiry was created if nobody ever has: COALESCE(max(note.created_at), deal.created_at).
An enquiry can be updated by a machine and still be abandoned by a human. A vector sync, a score being written, a stage tidy-up — all of those move updated_at and none of them is contact. The note is the human's fingerprint, and it is the only honest measure of attention. Same rule as Working the old list, and the year after the sale, where it decides who counts as an old lead.
Quiet means 7 days with no note. Serious means 21 days. Those two numbers are the entire policy, and they are the first thing to discuss with a client — not the automation, the numbers.
One query in, many tasks out
Everything so far has done one thing to one record. This does one thing to fifty, through a Loop — and the loop is the first node in this course with two different outputs that mean different things:
| Handle | When it runs | What hangs off it here |
|---|---|---|
loop-body | Once per row | The chase task for that enquiry |
loop-complete | Once, after the last row | The second query, and the escalation |
loop-body
- When it runs
- Once per row
- What hangs off it here
- The chase task for that enquiry
loop-complete
- When it runs
- Once, after the last row
- What hangs off it here
- The second query, and the escalation
So the escalation is not raised fifty times. It is raised once, after every individual task has been created, from a second query that asks a different question of the whole pipeline at once.
It stops at fifty, and it stops at the right fifty
Both the query and the loop cap out at 50. If a business has two hundred quiet enquiries on a Monday, a hundred and fifty of them are not chased that day.
What makes that acceptable is the ORDER BY above it: the biggest cases first. A cap with an order is a policy — chase the fifty that matter most, every day, until the backlog clears. The same cap without an order would be a bug that looks like a feature, quietly ignoring a different arbitrary hundred and fifty each morning.
Ask this of every workflow that processes a list. Where does it stop, and does it stop somewhere defensible? It is the question that separates an automation you can put in front of a client from one you cannot.
The arm that goes nowhere, on purpose
The second query counts the enquiries silent for 21 days and adds up what they are worth. Then a condition asks whether that count is above zero, and only the true arm is connected to anything.
Nothing is missing. If nothing is that badly stuck, the correct behaviour is to do nothing — no task, no mail, no “all clear” notification for somebody to start ignoring. You met this shape in Conditions and branches: an unwired arm is sometimes a guard, and reading which is which is a skill.
What the tasks actually say
Open the two writes and read them as an advisor would. The daily one refuses the obvious message; the escalation refuses the comfortable conclusion:
- Do not send a just-checking message — bring something new, or ask the honest question: is something in the way?
- A case with no owner and no date is the state cases die in, and it flatters the pipeline report while it does.
That is sales management, written into a workflow. The automation's contribution is the list of names; the value is in what the person is told to do with it. When a client asks what makes these workflows different from a reminder, this is the answer, and it is readable in a text box.
Change one thing — find your blast radius before you switch it on
Do not activate this one. You were told why in Draft and active: which workflows are actually running, and now you can see it: switching this on raises up to 50 tasks at 09:30 tomorrow. Find out how many before you decide.
- Open the first query and read it until you can say, in one sentence, who it returns.
- Run just that node and count the rows. That is exactly how many tasks tomorrow morning would bring.
- Change
7 daysto a number you would defend to this business, and run it again. Watch the count move. - Put it back, save, and leave the workflow as a draft.
Try it
- Work out what happens to an enquiry that is chased by this workflow every single morning for a fortnight, and decide whether you would change anything.
- Say what
continue_on_erroron the loop protects you from, and what it hides.
Next: Telling Meta which leads were worth it — the last two workflows in the catalogue, and the only ones that send anything back out of the org.
Related lessons
Base rates — what a piece of evidence is actually worth
A face-recognition system that is 99.9% accurate and almost entirely wrong, and a number that sent an innocent woman to prison. Both are the same arithmetic, and it is the arithmetic that decides what any piece of evidence is worth.
ReadConfirmation and survivorship — what you never looked for
Two questions about evidence you did not go looking for. One is a rule you have to discover, and one is a pattern in five famous people — and in both, the thing that would have told you the truth is the thing nobody checks.
ReadLoss aversion, sunk cost and regression — what it costs you
Four questions you answer about yourself rather than about a scenario, and your own answers are the finding. Then the pattern that makes praise look useless and criticism look like it works, whatever you actually do.
Read
