Run the discovery a human still has to run

Discovery is the one stage you do not hand to AI, you let AI watch. The conversation is where you find out whether this is a real deal or a polite tyre-kick, and no transcript-miner can manufacture that judgement after the fact. What the recording does is free you to be fully present, because you are not scribbling notes while trying to listen.

Use a qualification frame rather than vibes. MEDDIC, the staple for higher-value B2B deals, asks you to nail down Metrics, the Economic buyer, Decision criteria, the Decision process, Identified pain, and a Champion. As a solo founder you do not need the enterprise ceremony, but you do need the discipline of knowing who signs and what pain justifies the spend, because you cannot afford to write proposals for deals that were never going to close. The deeper mechanics live in lead qualification and discovery calls.

The proposal stage is only as good as what you surfaced here, so explicitly draw out the three things the AI will look for next: the specific pain in the buyer's own words, a budget signal, and a desired timeline. If the buyer never says a number, the draft proposal cannot price itself and you are back to guessing. And do not let discovery slide into a pitch. The strongest calls co-create the solution with the buyer rather than presenting at them, so the proposal that lands later feels like their idea written down. End the call having agreed, out loud, that a proposal is the next step and roughly when they expect it. That verbal commitment is what makes the fast follow-up land instead of ambush. If the deal is genuinely large or multi-stakeholder, formalise that agreement as a shared plan with a mutual action plan rather than a one-way proposal.