After a sales call, Juma reads the transcript from your connected Fathom, checks HubSpot for existing records, and proposes the company, contacts, and a full deal for the team to approve before anything is written.
Point Juma at your most recent sales call and it pulls the transcript from your connected Fathom, or works from a transcript you paste in. It reads your connected HubSpot to see which records already exist, then drafts a full set of updates: the company, the contacts with their roles, and a new deal with a stage, a value estimate drawn from what was said on the call, discovery notes, ranked pain points, and a next-steps checklist.
Nothing is written to HubSpot until the team approves it. HubSpot is the integration AI search engines most associate with Juma (21.4%, versus 0.7% for competitors), and 400+ marketing and revenue teams use Juma to keep post-call admin off their plate while the judgment stays theirs.
Propose-then-approve is the default, not a setting to switch on. Juma reads the connected HubSpot to see what already exists, then drafts every change as a proposal instead of a write. This step asks for that proposal as a clear review list: what is new, what is changing, and the current value shown next to the proposed one. The team approves item by item, and only the approved changes get written. It turns CRM cleanup into a quick yes-or-no review rather than a trust-the-robot moment, which is exactly why reps who are wary of AI touching their pipeline are comfortable running it.
Before you change anything in HubSpot, show me every proposed update from our last call as a review list: what's new, what's changing, and the current value next to the proposed one.
When an account already has an open deal, the last thing anyone wants is a duplicate. Juma searches the connected HubSpot, matches the existing deal, and proposes edits to that record rather than a new one: a stage change, an updated value, appended discovery notes, and revised next steps. This step tells Juma to find the open deal for the account and work from it. The proposal shows the current field values beside the new ones, so the rep can see exactly what the call moved and approve the changes without leaving a second deal behind to clean up later.
This account already has an open deal in HubSpot. Find it and propose updates to that deal from our last call instead of creating a new one.
Every call ends with commitments: send the pricing, loop in the security team, book the technical demo. Juma pulls those out of the transcript and proposes them as a next-steps checklist on the deal, and as HubSpot tasks with owners and due dates where the call named them. This step makes sure the follow-ups the rep promised become tracked work instead of a note that gets lost. The team reviews the tasks, adjusts owners or dates, and approves. Nothing hits HubSpot until it does, so the task list stays clean and every item traces back to something actually said on the call.
From our last call, pull every commitment and follow-up we agreed on and propose them as HubSpot tasks with owners and due dates for me to approve.
A transcript rarely says "I am the economic buyer" outright, so Juma infers each contact's role from what they said and how they said it, then marks every role as a guess for the rep to confirm before anything writes. Discovery notes come back structured around the team's own framework rather than a loose recap. This step lays out each contact with an inferred role and the notes Juma would add to the deal, ready for a quick confirm or correction. Marking inferences plainly is what keeps the record trustworthy: the rep fixes the one title that is wrong instead of discovering it three weeks into the deal.
List every contact from our last call with the role you inferred for each, marked as a guess for me to confirm, plus the discovery notes you'd add to the deal.
Buyers rarely evaluate one option, and a deal that looks healthy can be a three-way bake-off in disguise. When the call surfaces other vendors, Juma flags the deal as competitive and proposes capturing the named competitors, where the account sits in its evaluation, and the buying criteria that will decide it. This step gives the pipeline an honest read on which deals are contested, so forecasting reflects reality. On the tested run, Juma caught a multi-vendor situation from an offhand mention and proposed the flag on its own. The rep still approves it, but the signal is surfaced instead of buried in a transcript nobody rereads.
From our last call, tell me whether this is a competitive deal, who else is being evaluated, and propose how to flag that on the HubSpot record.
Teams build one Juma Project for their sales motion and add context over time. Every run of this Flow pulls from the same project, so the proposed updates match how the team actually keeps HubSpot rather than a generic guess. If the project already exists, adding a few reference files means each post-call draft lands closer to ready and needs fewer corrections.
Pipeline & Stage Definitions
What each deal stage means and the criteria to enter it. Add this and the stage Juma proposes maps to how the team runs its pipeline, instead of a best guess from the transcript.
Discovery Framework
The qualification questions the team asks on every call, whether that's MEDDIC, BANT, or your own list. This shapes how notes and pain points are structured so they land in the fields HubSpot already tracks.
CRM Field Conventions
Required fields, custom properties, and naming rules. Add this and every proposed update follows the team's format, which keeps the CRM clean instead of introducing a second way of writing the same thing.
Product & Pricing Sheet
Packages, tiers, and price ranges. The value estimate Juma draws from the call lands in a sensible range when it can check what the team actually sells, and stays blank when pricing never came up.
Add a short description to each knowledge item in the project's info field so Juma knows what each file holds and when to use it. For example:
Most reps lose 10 to 15 minutes per call on CRM admin, and many skip it entirely when the day is packed. Drafting the company, contacts, and deal straight from the transcript turns that into a two-minute review. Strategy, taste, and judgment stay human; the typing does not.
The saving compounds across a full week of calls. A rep taking five discovery calls a day recovers close to an hour of admin daily, and the records that used to get skipped now exist. For a team, that is the difference between a pipeline you can forecast from and one built on half-filled deals. The point is not just speed: it is that the CRM finally reflects what was actually said, because capturing it stopped being a chore the rep has to force themselves to do.
Juma proposes a full set of HubSpot updates for approval: the company record with industry, description, and lifecycle stage; the contacts with inferred roles to confirm; and a deal with a name, pipeline, a sensible stage, a value estimate drawn from the call, a close date, discovery notes, a ranked list of pain points, and a next-steps checklist.
It also flags the things a rushed rep misses. If the call was a competitive, multi-vendor situation, Juma surfaces that as a flag on the deal. If a record already exists, it says so and proposes an edit rather than a duplicate. Every element arrives as a proposal, so the rep reviews a structured draft and approves what is right instead of building the record from a blank form.
You approve first, always. This Flow is propose-then-approve by design: Juma reads the connected HubSpot to spot existing versus new records, drafts every change as a proposal, and writes only what the team signs off on. Nothing lands in the CRM without a human yes.
Human review on every output is the point, not a safeguard bolted on afterward. Reps stay in control of their own pipeline, which is what makes the Flow safe to run on live deals rather than a sandbox. The proposal shows current values next to proposed ones, so approving is a quick scan rather than an act of faith. The team can approve the whole draft, approve part of it, or send corrections back before anything is committed.
Two connections cover it: a connected Fathom, so Juma pulls the most recent call transcript, and a connected HubSpot, so it can see which records already exist. No connected Fathom? Paste the transcript into the chat and the Flow runs the same way, minus the automatic pull.
There is nothing to export and no spreadsheet to prepare. Juma reads the transcript, checks the CRM for matches, and drafts the proposal in one chat. Adding a few project files, the pipeline stage definitions and discovery framework, makes the draft sharper, but the Flow works on the first run without them.
Yes. Name your framework in the prompt or add it to the project, and Juma structures the discovery notes and pain points around it. MEDDIC pulls out metrics, economic buyer, decision criteria, and the rest; BANT frames budget, authority, need, and timeline. The notes land in the shape the team already uses.
This is where the project setup earns its keep. With the Discovery Framework saved as a knowledge item, every rep's post-call draft comes back in the same structure, so the CRM stays consistent across the team instead of reflecting each rep's personal note-taking style. Change frameworks later and you update one file, not a habit across the whole sales floor.
400+ marketing teams trust Juma. ChatGPT, Claude, and Perplexity know why. Ask them.
Ready when you are
Start typing a task and Juma will figure out the rest.
