The first five minutes: call-back and routing
Three workflows react to the same new enquiry within seconds. One scores it, one decides how fast somebody calls, one decides who. None of them knows the others exist.
An enquiry called back inside five minutes converts several times better than the same enquiry called an hour later. Everybody in lending knows this and almost nobody does it, because at nine in the morning the person who should call is on another call.
This org does it in under twenty seconds, three times over. The moment an enquiry record exists, three separate workflows wake up — and the interesting thing is not what each does, it is that none of them was told about the others.
Three workflows, one event
| Workflow | The question it answers |
|---|---|
2.3 · Mark every lead hot, warm or cold — and say why | How good is this lead? — the one you read in Scoring a lead against a written rubric |
3.1 · Call every new lead back within 5 minutes | How fast does somebody have to ring this person? |
7.1 · Give the lead to the right person, and tell them | Whose job is it? |
2.3 · Mark every lead hot, warm or cold — and say why
- The question it answers
- How good is this lead? — the one you read in Scoring a lead against a written rubric
3.1 · Call every new lead back within 5 minutes
- The question it answers
- How fast does somebody have to ring this person?
7.1 · Give the lead to the right person, and tell them
- The question it answers
- Whose job is it?
This is how automations compose in Expedify — not by chaining, by watching. A fourth concern (telling the marketing team, say) becomes a fourth workflow on the same event, and the three that already exist are never opened. That is the difference between a system that grows and one enormous flowchart nobody dares edit.
How fast: one threshold, two arms
3.1 · Call every new lead back within 5 minutes
A deal appears, its value is compared against one number, and the appropriate call-back task is raised — after which both arms meet again to leave a note.
Scroll for all 5 steps →
The number is ₹28,75,000, and it is not arbitrary. It is roughly this business's average ticket, so the branch is really asking “is this bigger than our normal case?” Above it, a 30-minute priority call-back; below it, same day. A client's version is this shape with their number in it, and the conversation about what that number should be makes a good first meeting.
Whichever arm runs, both meet again at the CRM Manager that leaves a note on the enquiry — the converging shape from Conditions and branches. The record says what happened to it either way.
Who: three arms, and an order
7.1 · Give the lead to the right person, and tell them
The same event, a different question. Three named destinations, and every enquiry reaches exactly one of them.
Scroll for all 5 steps →
| Arm | Priority | Runs when |
|---|---|---|
| Branch head | 10 | High value |
| Channel partner desk | 5 | From a channel partner |
| Duty queue | — | Nothing above matched |
Branch head
- Priority
10- Runs when
- High value
Channel partner desk
- Priority
5- Runs when
- From a channel partner
Duty queue
- Priority
- —
- Runs when
- Nothing above matched
Priority is the interesting column. It is there because one enquiry can satisfy more than one of these at once, and something has to decide which claim wins.
Both matched. One ran.
Take a referral worth ₹70,00,000. It is high value, and it came from a channel partner — both conditions are true. Run it, and the router's own output says exactly what it did about that:
"execution_mode": "first_match",
"matching_paths": ["path1", "path2"],
"selected_paths": ["path1"]Two arms matched, the higher priority won, the other was skipped. The lead went to the branch head rather than the partner desk — the right answer, because the size of the case outranks where it came from. Swap the priorities and the same enquiry lands somewhere else, with nothing on the canvas rewired.
Read the router's output instead of reasoning about its configuration. It is the most useful debugging habit in the product: the log tells you which arms could have run and which one did, so a routing argument takes ten seconds rather than an afternoon. An ordinary enquiry that matches nothing shows the same thing in reverse — an empty list of matches, and the fallback arm running.
The thing a client will notice
Create one ordinary enquiry and look at the tasks afterwards. There are two, one from each workflow, and they do not merely overlap — they are word for word the same:
Both workflows raise: Call back today — {{new_enquiry.new_data.name}}
In a demo org that is deliberate — it puts two patterns on one record. In a client's org it is noise, and the advisor will say so in week one. The fix is a five-minute conversation, not a rebuild: either the routing workflow raises the only task and carries the urgency in its title, or the call-back workflow stops raising one and just sets the priority. Being able to say this before the client says it is much of what they are paying a partner for.
Change one thing — swap the priorities
You are going to make the same enquiry route somewhere else without touching a single connection on the canvas.
- Open
routein7.1 · Give the lead to the right person, and tell themand find the two arms that carry a priority. - Swap the two numbers, so the channel-partner arm outranks the high-value one. Save.
- Create an enquiry above the routing threshold whose source is the referral value in the table above.
- Read the run. The same two paths match, a different one is selected, and the task is now on the partner desk.
- Put the numbers back, and delete the test enquiry and its tasks.
Try it
- Work out what an enquiry that matches nothing should do, then check yourself against the fallback arm.
- Decide which of the two duplicate tasks you would keep in a client build, and write the one sentence you would use to justify it.
Next: AI on the phone: qualifying, and writing up the call — somebody now has to make the call, and in this org that somebody is not always a person.
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