Update your CRM from call transcripts
Read new call transcripts and draft CRM updates for you to approve in Slack.
CRM automation fixes the oldest problem in sales: the call happens, the notes don't. This workflow runs the moment Fireflies or Granola finishes a transcript, extracts what changed (stage, next steps, objections, new contacts) and updates the deal in Attio or HubSpot. You approve the changes in Slack before they're written.
What does this workflow do?
After every recorded sales call, the agent reads the transcript and pulls out the facts a CRM needs: deal stage signals, budget and timeline mentions, objections raised, competitors named, new stakeholders on the call, and the agreed next step.
It then drafts the CRM update: field changes, a call summary on the deal timeline, new contacts to create. Nothing writes automatically unless you want it to. The default is a short approval message in Slack: here's what I heard, here's what I'll change, approve or edit. One reply and the CRM is current.
How does it work?
- A finished transcript starts the run. The workflow runs the moment Fireflies or Granola finishes a call transcript. You can also run it on demand for a specific call.
- The agent matches the call to a deal. It finds the right record in Attio or HubSpot from the attendees and account. If it can't match confidently, it asks you.
- It extracts structured updates. Stage, close date signals, objections, competitors, stakeholders, next steps. It knows your pipeline stages and your field conventions because it has ingested your CRM setup.
- You approve in Slack. A compact diff of proposed changes; reply in Slack to approve or correct it. Reps who trust it can switch to auto-write per field.
- The CRM gets updated. Fields change, the summary posts to the timeline, and follow-up tasks are created.
Why automate CRM entry at all?
Reps skip CRM hygiene under deadline pressure, and forecasts inherit the gaps. Automated data entry keeps the pipeline honest without adding admin time to anyone's day.
The workflow also improves with use. When a rep corrects an extraction ("that wasn't an objection, that was a joke"), the correction persists. When your team renames a stage or adds a field, the agent picks it up from the CRM. And because Agentwork holds your company's knowledge, the call summary can note things a transcript tool can't: that this prospect also talked to your team eight months ago, or that the competitor they mentioned is one you have a battlecard for.
Works with
Fireflies, Granola, Attio, HubSpot, Slack. Runs when a transcript finishes, or on demand for a specific call.
Frequently asked questions
Does it write to the CRM without asking?
Not by default. The default flow sends a proposed diff to Slack, and a quick reply approves it. You can enable auto-write per field once you trust the extractions, and keep approval on sensitive fields like deal stage.
Which call recorders does it support?
Fireflies and Granola today. The workflow reads finished transcripts, so any call your recorder captures is in scope.
What if the agent matches a call to the wrong deal?
Matching uses attendees, email domains and account names. When confidence is low, the agent asks the rep to confirm the deal before anything is written. Corrections are remembered.
Can it create new contacts and companies?
Yes. New stakeholders who appear on calls are proposed as contacts, linked to the account, with their role noted from the conversation.
How is this different from HubSpot's own call logging?
HubSpot logs that a call happened and attaches a recording. This workflow turns the content of the call into field updates, contacts and next steps, with your approval, and carries context from your company's other tools into the summary.