HubSpot's Deal Progression reads calls and emails and suggests CRM updates a rep approves with a click. Use it for notes and drafts. Keep stage, amount, close date and next step with the rep.
HubSpot Deal Progression, launched at UNBOUND in September and in public beta, reads emails, calls and meetings, suggests next actions, drafts follow-ups and proposes CRM updates. It's useful for notes and drafts. Keep deal stage, amount, close date and next step as rep decisions, ask for a one-line reason on stage changes, and review rejected suggestions weekly.
At its UNBOUND event in September, HubSpot launched Deal Progression. HubSpot says it "analyzes every email, call, and meeting to surface recommended actions, draft follow-ups, and update your CRM automatically." A partner's rundown of the launch describes it as proposing property updates that the rep approves with a click. It's in public beta.
I like the idea. Reps hate CRM admin, and most CRMs are out of date because of it. But "approve with a click" has a catch, and sales managers should think about it before rolling this out.
When a rep updates a deal by hand, they have to decide something. Is this deal really in Proposal? Is the close date still the 30th? Did the buyer actually agree to a next meeting, or did they say "let's talk soon"?
That decision is the useful part of CRM work. The typing never mattered.
With Deal Progression, the AI makes the first call and the rep confirms it. That's faster. It also means the rep's judgment shrinks to a glance at a suggestion. If the suggestion is wrong and looks reasonable, it goes through.
Approving a draft is a different skill from writing one. People get quick at it, and then they get careless at it.
Picture a typical call. The buyer says, "This looks great, I'll take it to my boss next week." An AI listening to that will likely see positive sentiment and a next step. It might suggest moving the deal forward and setting a follow-up for next week.
An experienced rep hears something else. "I'll take it to my boss" can mean the buyer has no authority. "Next week" with no day isn't a next step. The deal may be stuck, not moving.
These are the places where call transcripts and real deal status often split:
None of this means the tool is bad. It means the AI reads what was said. The rep has to know what it meant.
HubSpot's own launch page quotes a customer who says they approve about 73% of what Deal Progression suggests, with next steps fully automated, and that they are "earning the right to automate more by letting the signal prove itself." That's a sensible way to think about it. Start narrow, check the results, then widen.
Here's how I'd split the fields on day one.
Fine to accept with a quick look:
The rep decides these:
The second list is your forecast. If those fields drift, your pipeline report becomes fiction. Keep them human for at least the first quarter, and look at how often the AI's suggestions would have been right before you hand any of them over.
One small rule helps a lot. Any time a deal changes stage, the rep writes one line on why, in their own words.
"Moved to Proposal: buyer asked for pricing for 40 seats and named the CFO as signer."
That line takes ten seconds. It forces the rep to check the suggestion against what they actually know. And it gives the manager something to read in a pipeline review besides a column of stages.
If the rep can't write the line, the deal shouldn't move.
For the first few weeks, add one question to your weekly pipeline meeting: which AI suggestions did you reject, and why?
The answers tell you a lot. If reps reject almost nothing, either the tool is very good or nobody is reading. If they reject a lot on one field, that field isn't ready to automate. If one rep rejects much more than the others, they might know their deals better, or they might not trust the tool. Either way, you want to know.
The part of Deal Progression I'd turn on first is the follow-up drafts. A draft email that the rep edits before sending is low risk. The buyer only sees what the rep chose to send.
The draft still needs editing. AI-written follow-ups tend to be long and polite and say little. Cut it to one fact the buyer said, one question they can answer, and a sign-off. If the draft makes that faster, it's worth it.
HubSpot pitches this as a CRM that stays up to date without manual entry. That's a real benefit. But a CRM full of fresh, confident and wrong data is worse than one that's a bit stale. Stale data at least looks stale.
The point of keeping the CRM current was always to know which deals will close. If the tool saves your reps a few hours a week and they spend some of it thinking harder about the deals that matter, great. If it saves them those hours and they stop thinking about deal status at all, your forecast will tell you in about a quarter.
Turn on the drafts. Keep stage, amount, close date and next step with the rep. Read the rejections. Widen the AI's reach only when the numbers say it's earned it.
SalesAssist installs the complete follow-up and outreach systems behind 7-figure businesses — so your reps spend 100% of their time closing.
It's a Sales Hub feature HubSpot launched at UNBOUND in September 2026. It reads a deal's emails, calls and meetings, recommends next actions, drafts follow-ups and proposes CRM updates for the rep to approve. It's in public beta.
Salesforce's Hunter is an outbound sales agent built to work deals over weeks, with GA listed for November. Set the guardrails before it starts making a dead deal look alive.
"Follow up" isn't a next step. If the buyer hasn't agreed to a date, the deal belongs in your hopes, not your forecast.
The email that gets a reply is shorter than the CRM template. One fact from their mouth, one question they can answer, then stop.