Renewal Hawk counts what your team actually writes into the CRM, not what they were licensed to write. Seats nobody touches get cancelled. Seats that exist only to type into fields get taken over — the person keeps doing the work from Telegram, WhatsApp or a voice note, and the licence goes back.
Connect HubSpot read-only. Renewal Hawk counts every contact, company and deal update, every note, task, logged call, meeting and email — attributed per user, over 90 days — and sorts your roster into four buckets: active, light, delegable, dormant. Then it does the arithmetic against what you actually pay.
A report alone is just an opinion with a number on it. Which is why there is a second half.
The delegable bucket is the interesting one: people who open the CRM twice a month to change a field, leave a note, close a task. Renewal Hawk takes that over. They message it — typed or spoken — it finds the record, shows exactly what it is about to change, and writes only after they say yes.
They keep the job. You stop paying a full seat for the window they did it in.
Live output from the scan engine, running against a demo tenant. Point it at your own HubSpot and the shape of your roster replaces this one.
| Person | Role | Writes / 90d | Active days | Delegable | Verdict |
|---|
What this cannot see. Write activity is a floor on usage, not a ceiling: reading a dashboard, building a report or sitting in a sequence editor leaves no API trace. So no seat is ever cancelled on this number alone — every candidate goes through a 30-day shadow period first. A finding you cannot defend in the room is worth nothing.
The default share is what the demo tenant measured: 6 dormant, 9 delegable and 11 unaccounted out of 42. In the audits we have run, under 20% is rare.
Everything below happens through the CRM's own API, attributed to the person's owner record, and lands in an audit log you can export. Try it — the sandbox reasons and searches; it writes only when a HubSpot token is connected.
Live today. Inline yes/no on every write, voice notes transcribed, one bot per tenant.
Cloud API integration is written and waiting on number verification. Same gate, same audit trail.
Speak it walking out of the meeting. Gemini transcribes, the agent proposes, you tap yes.
The channel is a thin adapter over one runner. Slack, Teams or email is a day, not a quarter.
Zero CRM writes in the window. Shadow for 30 days, then cancel at renewal.
At least 85% of writes are record updates, notes, tasks or logged calls, at 25 writes/month or fewer. An agent can do all of it.
Under 4 writes/month but some of it is meetings or email sent from the product. Downgrade the tier; do not cancel.
Everything else. Keep the seat and stop talking about it.
Billed seats with no matching user at all. Leavers, duplicates, project accounts. The cheapest money on the table.
Applied to the capability "keep CRM records current" — the framework from Navigating B2B Sales — The Agentic Age Unpacked.
Low. A wrong field is visible, versioned, reversible. Nobody signs a contract on this.
The vendor's. Keep buying the data model, integrations and compliance. None of it is being rebuilt.
High. Your stages and properties are yours. A generic UI charges a full seat to express them.
Low. CRM object APIs are versioned and move slowly. Safe to own a thin layer on top.
Contained. Scoped writes, human confirmation, full audit log. Worst case is one bad field.
Verdict HYBRID. Keep or Buy the CRM. Build the window you look at it through.
Read-only HubSpot connection, seat verdicts, savings model, one-page report. Yours whether or not you go further.
The agent takes the seat's job. Telegram and voice today, WhatsApp on verification. Confirmation gate and audit log included.
The whole GTM stack, not just the CRM. Five Tests per capability, 3-year TCO, board-ready readout.
Point Renewal Hawk at the CRM contract that renews next. If it finds nothing, you have lost an afternoon and gained an argument for the vendor.