The call was good. You hung up, the recording finished processing, and you have forty-five minutes before the next one.

In those forty-five minutes, here is what usually happens: the transcript sits in a tab you will forget, the CRM record stays as it was, a Slack message asks "did anyone take notes?", an overeager agent or Zapier step tries to create a contact from the misspelled name in the transcript, and the POC task — the whole point of the call — does not exist anywhere a human agreed to.

By Friday, you have a partial HubSpot record, a duplicate contact, no POC scope document, and a follow-up email that references the wrong company name.

For a 2–25 person consulting firm, this is not a transcription quality problem. Transcription will keep misspelling names. This is a structure problem: there is no place where the call lands as a first-class object, no gate that holds entity creates until a human confirms, and no start signal that means "this POC is authorized to begin."

Cockpit closes that loop — and Cut A is the captioned walkthrough of the full path.

Why the post-call gap compounds

Every discovery call a small consulting firm takes is evidence. Evidence that a buyer has a problem, that you heard it, that there is a deal worth pursuing. That evidence evaporates if it does not land in a structured form within the hour.

The tools that exist do not help cleanly:

• Meeting recorders give you a transcript file. They do not give you a matched contact, a deal, or a task. They give you text.
• CRM systems have import tools. Import tools create duplicates. Ask anyone who has run a "merge weekend" before a board deck.
• Agents — Cursor, Claude, any of them — can read the transcript and propose a deal record. Without a gate, they will. And they will spell the buyer's name the way the ASR did, not the way the buyer does. You will spend twenty minutes cleaning the record the agent helpfully created.
• Asana or project tools can hold a POC task. But someone has to create it, link it to the right account, and make sure it has scope. That someone is usually you, after two more calls, when you have no context left.

The gap is not missing automation. The gap is missing structure: a place where the meeting is an object, entity creation requires confirmation, and work start requires approval.

What Cockpit does on the meeting-to-deal path

When a discovery call ends and the recording arrives in Cockpit, the sequence is:

1. Meeting ingest. The recording and transcript are a first-class intake card — not a file in a download folder, not a paste into a Slack thread. The meeting is the starting artifact, and everything downstream traces back to it.

2. Entity match, not entity create. Cockpit does not write a CRM record from the misspelled ASR output. It runs entity match search against the people and companies you already have, surfaces near-matches (including typo-tolerant hits), and queues a human confirmation in the approvals inbox: same person, new person at a known company, or net-new? Only after the confirm does the record stick.

   This is the same gate the entity match how-to post covers in detail: search without a gate still invents rows when confidence is soft; a gate without search dumps every misspelled name on you as a blank form. Together, the system proposes candidates and you decide.

3. Approvals inbox. The propose-create-deal and propose-start-POC cards land in one inbox alongside every other decision the system surfaces. Agents propose. You swipe. Nothing starts before you say go.

4. Approve-before-start. The POC task does not exist until you approve it. The deal record does not update until you confirm the contact. There is no auto-start. Cut A shows this beat by beat — the desk proposes two cards (Create Deal, Create POC Task) and the agent waits until both turn green.

After both approvals, the scoped work runs with the meeting recording in context. Done checklist in view. The discovery call never leaves the work item.

The proof trail that matters for proposals

For a consulting principal, the meeting-to-POC path matters beyond delivery — it is the proof trail that lets you quote a client accurately and defend the scope later.

When you run a proposal six weeks after a discovery call and the contact record is a duplicate, the deal stage is blank, and the POC never had a task with acceptance criteria, you are reconstructing scope from memory and a transcript that spells the client's name wrong three different ways. That is how scope creep starts and why flat-fee projects go underwater.

The Cockpit path produces a durable record:

• The meeting as an intake object, linked to the transcript
• The contact and account confirmed by a human before they stuck
• The POC task with scope created after approval, not before
• The deal record that traces back to the call that originated it

That record is the operating asset. Not a CRM dashboard screenshot — the actual sequence of decisions that produced the scope.

What is shipped (and what is not)

Fair to name today: meeting ingest (call as first-class object), entity match including typo-tolerant search, the approvals inbox for contact/deal/task-start confirmations, and approve-before-start as the gate before POC work begins.

Not shipped: hooks and execution logs tied to live agent sessions (labeled coming soon on the product page), streaming agent UI inside Cockpit, Claude CCMA as a third runtime option, and client handoff embeds. Those are on the roadmap.

The underlying infrastructure is @Supabase as the headless DB — agents, CRM, and project management in one schema — and @vercel as the UI layer. Cockpit is the control plane on top of that stack, not a replacement for either.

The consulting firm that runs it on their own shop first

The pitch for small US consultancies is not "another tool." It is the implement-for-others motion: buy the AI (~$300–500/mo for Cursor, Claude, Grok, the host stack), run Cockpit on your own work, and when a client asks "can you set this up for us?" the answer is "yes, and here is what we learned."

The discovery-to-POC path is the first place to run it, because it is the path principals take multiple times a week. Every call that lands clean — matched contact, approved deal, scoped POC, proof trail — is a rep in the methodology you will sell.

See the full Cut A walkthrough at /blog/how-to/cockpit-cut-a-meet-recording-to-poc-approve-before-start. See the product at cloudbeast.io/cockpit. Book a discovery call at cloudbeast.io/schedule to talk through what the first desk looks like for your practice.