Turn on lead capture
Actions on the agent is where this lives, and the setup checklist links straight to it. Once it is on, a conversation that shows buying intent ends with the agent asking for a name and a way to reach them, rather than with a polite goodbye.
It is the single highest-return switch in the product for anyone whose website chat is a sales surface as well as a support one. The questions people ask before buying are the same ones they ask after, and only one of those conversations is worth a follow-up call.
Agents → Actions
Turn on lead capture. It is off on a new agent.
Decide what you need
A name and one contact method is usually enough. Every extra field is a chance for them to stop.
Watch the Inbox
Leads appear as they arrive, with the conversation they came from attached.
Taking a message
Separately from leads, every agent can take a structured message — the thing it does when a customer needs something it should not attempt, and no person is available to take over. Those land in the Inbox too, and with SMS notifications configured somebody is told rather than finding it the next morning.
This is what makes out-of-hours handover honest. Instead of promising a transfer to a person who is not there, the agent takes the details and says when somebody will be in touch.
Reading the results
The Inbox has two lists: leads, and the extra fields collected in conversations. Every entry links back to the conversation it came from, which is usually the thing that tells you whether it is worth a call — a name with the question that produced it is a far better lead than a name on its own.
Contacts is the same people, deduplicated across channels. Somebody who chatted on Monday, emailed on Wednesday and called on Friday is one contact with three conversations, not three strangers.
Getting it out of WireDesk
Leads, collected fields, contacts and conversations all export as CSV, which is deliberately the boring answer: a CSV opens in the spreadsheet your team already uses and imports into the CRM you already pay for, with no integration to maintain.
For deeper wiring, actions can call your own systems — checking an order, booking a slot, looking something up — so the agent answers from live data rather than from a page about it. That is the point where this stops being a chatbot, and it is worth a conversation with us before you build it.