Free workshop

1:1s that prepare themselves

A badly prepared 1:1 feels like a good conversation: you talk, you listen, everyone leaves happy. Three weeks later nobody can say what was decided. The cause isn't your memory: it's that your meetings produce nothing a machine can re-read. Flip the order — raw material first, intelligence second.

20 min to set up · then 2 min before each check-in

Free · no sign-up · prompts included

Before you start Set up your AI in 20 minutes

How we do it

  1. Transcribe everything, starting now

    No exceptions: 1:1s, team meetings, video calls and in-person. What isn't transcribed won't exist in the context of your work. And you never miss information you don't know you lost.

  2. Separate capture from understanding

    Your AI doesn't transcribe: it joins no call and hears nothing. One tool captures, the AI reads, files and uses. Don't try to make one piece of software do both.

  3. One folder where everything lands, raw

    That's deliberate: you should never have to decide, mid-flow, where a write-up goes. Filing is the agent's job, not yours. Alongside it, one folder per person and a follow-up folder.

  4. An agent that files every evening

    It's the piece everyone skips, and it's the one that keeps the rest alive. Its rules live in a file, not in the prompt: when a rule changes, you edit the document, never the task.

  5. Four statuses, one of them for doubt

    Filed, out of scope, needs routing, duplicate. Faced with an ambiguous case the agent doesn't guess: it leaves the file where it is and flags it. One bad automatic filing costs more than ten human calls.

The prompts, ready to copy

Prompt 1 · the filing agent

Set it up once, as a scheduled task.

Create a scheduled task “file the transcripts”, every day at 7pm.

Your rules live in transcript-rules.md at the root of the project. Re-read it on every run: that file is the authority, not these instructions.

On each run:
1. List the new files in the transcripts/ folder and compare them to the index at the bottom of transcript-rules.md. Only handle what isn't there yet.
2. For each one, work out what it is: a client (which), an internal project (which), a 1:1 (with whom), or something that doesn't belong here. Use the participants, the title and the content.
3. Rename it YYYY-MM-DD-type-counterpart.md, then move it to the right folder.
4. Add a line to the index: date, title, status, destination. Possible statuses: filed, out-of-scope, needs-routing, duplicate. Never delete an existing line.
5. If you hesitate, move nothing: status needs-routing, and flag it to me.
6. If a file already exists for the same date and counterpart, don't create a second one: append what the new one adds to the existing file, and mark it duplicate.
7. Finish with three lines: what you filed, what you left, and what deserves my attention in what was said.

You never delete a file. You contact nobody.

Prompt 2 · prepare the check-in

The one you'll use every week. Line 5 is the uncomfortable one, and the most valuable.

Prepare my 1:1 with [name].

Re-read the transcripts of their last two check-ins, their follow-up file, and what we've exchanged by email since. Then give me, in five lines maximum:
1. What was decided last time, and where it actually stands.
2. A question about progress — specific, one that calls for more than a yes.
3. A question about blockers, phrased so it's easy to answer honestly.
4. A commitment of mine to re-check: something I promised them.
5. A topic we've been pushing back for more than three check-ins, if there is one.

No introduction, no conclusion. The five lines, that's it.

The three mistakes that ruin it

  • Transcribing selectively If you pick and choose which meetings to record, your agent works on a memory hole. It's all or nothing.
  • Filing by hand Three days, then you give up, and the folder becomes a junk drawer of six hundred files. Scheduled task from day one.
  • Letting a follow-up go out unread The draft is good, it isn't accurate. You always approve. The day you stop reading them, you'll find out why you had to.

Transcribing meetings means processing personal data about your teams and your customers: tell them, check the legal basis, and set a retention period. And keep out of the shared repository what doesn't belong there: personal matters, other companies' business, anything that shouldn't circulate.

All workshops

Set up your AI in 20 minutes

The first move: give it your context once.

20 min · once · no technical skillsOpen the workshop

Emails written in your voice

It learns your style from your sent mail, and drafts your replies.

20 min for the style · 10 more for the automationOpen the workshop

Your company context, written down

What the AI needs to know about your company, written once and maintained.

20 min for v1 · then 15 min a weekOpen the workshop

Would you rather we did it with you?

These workshops are the starting point. On client work we do the same thing at company scale: your company's context written down properly, the processes automated, and your teams trained while we build.

Request your audit

30 minutes · no commitment