Skip to main content
Reps update CRM on Friday afternoon, from memory, for the deals they remember. Everything downstream inherits that: forecast, pipeline reports, the deal review that starts with someone reading a stage field nobody has touched in three weeks. CRM updates fills the fields from what actually happened. The call said the champion changed, so the contact role changes. The email said the timeline moved, so the close date moves. The rep does not retype what a system already knows.

​
Data Sources

Starter (required only): field updates drawn from what was said on calls, written back to the deal or account record. Full (required + recommended): the same updates, plus what changed over email and in Slack, so a timeline that moved in a thread is caught even when it never came up on a call. See the Context Graph for how each source feeds the update.

​
Choose your path

​
Build in Clearskies

Build the workflow once and CRM updates itself from the record of what happened. Start it after calls, or on a schedule that sweeps the deals that moved.

​
The Workflow

In the Clearskies app, open the workflow builder and describe what you want in plain language:
The app configures the trigger (meeting ended), the fields in scope, and whether writes land directly or wait for a person. Meeting workflows always start with a filter, so the workflow only runs on real customer calls and not on out-of-office blocks or internal syncs. Two things worth deciding up front. Which fields are in scope: start with two or three that are always stale and always matter, usually next step and close date, rather than every field at once. Whether a person approves: writes can go straight in, or pause for the deal owner to approve or reject. Most teams start with review on and turn it off field by field as they build trust. Test runs never write. You can point a draft at a real closed call and see exactly which field would change to what, with nothing touching your CRM.

​
Deploy to Slack

With review turned on, the deal owner gets the proposed write before it lands:
Approve and the fields update. Reject and nothing is written and the run stops.

​
Customize the rollout

  • Pick the fields in scope, and add to them as the team trusts the writes
  • Require approval on some fields and let others write straight through
  • Route approvals to the deal owner, or to a manager for larger deals
  • Run it on a schedule instead of per call, to sweep deals that moved over email
  • Chain it with post-call follow-through so the same call produces the recap and the field updates
Write-back is supported for Salesforce and HubSpot on connected integrations. HubSpot writes cover accounts, contacts, and deals. See Write to Salesforce and Write to HubSpot for what each field type accepts.

​
Build in Claude or ChatGPT

Useful before you automate anything. Ask what should change, read the answer, and decide which of those fields you trust a workflow to write.

​
The Prompt

Open Claude or ChatGPT with the Clearskies MCP connected and paste this (or modify to your needs). Replace the account name with your own.
The gap between what CRM says and what happened is usually the most useful output here, before any automation exists.

​
Make it yours

  • Add Do this for every open deal in my pipeline to size the hygiene problem across the team.
  • Add Only flag fields that affect the forecast to narrow it to what leadership reads.
  • Add Draft the CRM summary note I should paste into the opportunity when the field you care about is a text field.

​
The Plugin

Ready to make this part of your team’s workflow? We’ll set up the Clearskies plugin with you. Clearskies workflows, Claude skills, all tailored to how your team works. Book 15 minutes →

​
Next steps

  1. Sign in to Clearskies
  2. Connect your data sources
  3. Get your Clearskies MCP server and try with Claude or ChatGPT