The problem
Someone asks what a custom order would cost. Your agent answers what it knows — materials, lead time, the usual range — and then stops, because the actual number is yours to decide.
And that is where it goes wrong. The request is now one message inside one conversation, in a list of conversations. Nothing is holding it. You find it tomorrow, or on Sunday, or when the customer follows up somewhere else — and by then they have asked two other people the same question.
How it works
Three moving parts, and you only set up the first two once:
- Your agent recognises a quote request and puts a label on the conversation.
- Your helpdesk knows that label means a person has to act, so the conversation lands on a list.
- You work that list — read what they asked for, price it, send it.
Both halves, or neither. This is the one thing people get wrong. The agent has to apply the label AND the helpdesk has to be told what it means. Set up only the second and your list stays empty forever — which looks exactly like nobody asking for quotes.
Step 1 — teach the agent
First, the agent has to notice. You are describing a situation, not keywords — it reads the intent across the whole conversation, so "how much for 40 of these" and "can you price a custom set" both count.
Open Settings, then Tags in the side menu under Automations, and add a rule.
If you have connected Claude, just describe the rule and it sets it up. Same result, no screens.
Write the label down. You need it again in the next step, spelled exactly the same way, in the same language. This is the one place a typo costs you the whole feature.
Step 2 — tell the helpdesk
The agent is now labelling conversations. Nothing happens with them yet — a label on its own is just a label. Now you say what it means.
In your Helpdesk, open Settings → Labels that need a person and add one rule.
What has to happen is written for your team, not the customer. It is the line the list shows, so make it the actual next action.
Ask Claude for it in one sentence:
It reads back what it saved, and it checks the half that quietly fails without being asked.
Nothing changes for anyone else. Only labels you name here do anything. Every other tag your agent applies keeps behaving exactly as it does now.
Working the list
From here on it runs itself. A customer asks for a price, the agent tags the conversation, and it appears on your list with the line you wrote and how long they have been waiting.
Open your Helpdesk inbox and filter by the label. Everyone waiting for a price is there, oldest first — and the ones where the customer spoke last are the ones to do first.
Open one, read what they actually asked for in their own words, and reply. When you have sent the quote, close it.
This is where a connected Claude earns its keep, because the whole job happens in one conversation. Things worth asking:
- “Who is waiting for a price quote?”
- “What did the third one actually ask for?”
- “Take the bot off that conversation while I work out the price.”
- “Send them this quote.” — then paste or dictate it
- “Mark it done.”
You can prepare the quote in the same chat you send it from, which is the point: nothing gets copied between windows.
Two things to know
WhatsApp has a 24-hour window. You can send a normal message for 24 hours after the customer’s last one. After that, WhatsApp only allows an approved template — so for a quote that lands the next day, prepare a template with the price as a variable.
Sending a message quiets the agent on that conversation. As soon as anything goes out from the business, your agent stops replying there and it is yours. That is deliberate — you do not want it answering mid-negotiation — but it does mean the next question is yours to catch.
The same trick, other situations
Nothing here is about quotes. It is "when the agent sees this, put it on my list" — so any situation you keep losing works the same way. Add a rule per situation:
- Complaint — “Call them today”
- Bulk or wholesale enquiry — “Send trade pricing”
- Return or exchange — “Approve or decline the return”
- Asked for a human — “Call them back”
Start with one. One rule you actually work every morning beats six you stop looking at. Add the second once the first is a habit. And see Set up tags for how the agent decides what to label in the first place.