Where the human stays, and where the machine takes over

Open a blank document and write each step of your sales process, in order, as it happens today. Include the small ones: scheduling, taking notes, copying the pain into the CRM, writing the proposal, chasing it, creating the invoice. Most solo operators find 12 to 18 steps. Do not judge them yet. Just list them.

For each step ask: does this require reading a person, or moving structured data from one box to another?

• Reading a person means deciding whether the buyer is serious, whether the economic buyer is in the room, what price you will hold and which risk to name. These stay with you.
• Moving data means transcribing, extracting pain and budget, matching needs to packages, filling a template, logging the deal and scheduling the chase. These go to the machine.

When a step is half and half, split it. Drafting a proposal is data. Deciding what the proposal should promise is judgement.

Everything between these three can be plumbing:

1. The conversation itself. Be fully present and let a tool listen. See Run the discovery a human still has to run.
2. The qualification call: is this deal real? You decide, after the call, with the call notes in front of you.
3. The final edit of the proposal before it leaves your hands.

Do not delegate these, and do not let the tools quietly take them over. A founder who automates the relationship and hand-edits the plumbing has the whole thing backwards.

Build one piece at a time and let each run for a week before adding the next.

1. Recording and transcript. Set a notetaker such as Fireflies, Otter or Granola to join every external call. Record every discovery call by default covers consent and setup.
2. Extraction. After each call, pull out pain in the buyer's words, budget signal and timeline, and write them to the deal in your CRM, for example HubSpot.
3. Draft. Feed the transcript, your positioning and your service catalogue into a drafting step. Turn the transcript into a proposal draft shows the inputs.
4. Follow-up and signature. Reminders, tracked sends and what happens on signing. See Chase the follow-up the AI remembers for you and Let the signature trigger everything downstream.

Connect the steps with a no-code tool such as Zapier or Make so nothing is retyped.

The proposal review is your third checkpoint, so make it quick and consistent. Before anything leaves, check five things:

1. The economic buyer is named, and they are on the distribution list.
2. The price matches what you said on the call.
3. The scope matches what the buyer agreed, with nothing invented by the draft.
4. One real risk is named, along with how you handle it.
5. The next step has a date.

If a check fails, fix it by hand or change the prompt that produced the draft. Aim for a review of ten minutes or less.

• Letting the AI score or qualify deals and then trusting the score without listening to the call.
• Sending drafts without a human read. The first time it invents a package or a price, you will remember it.
• Automating before the process works by hand. If your steps are unclear, automation only makes the confusion faster.
• Adding five tools at once and not knowing which one broke.
• Skipping consent when you record. Say it out loud at the start of the call.

You can measure it. The time from the end of a call to a reviewed proposal falls from days to under an hour. Your hours on admin per deal fall, while your hours in conversation rise. Every proposal you send comes from a deal you judged to be real, and the deals you disqualified did so for reasons written down in the CRM. If the close rate dips, check the three human checkpoints before you blame the tools.