Made for Swiss self-employed & SMEs
The client call ends well. You scribble three lines — “send offer, clarify VAT, ask about the March deadline” — and get back to work. Ten days later the note resurfaces under an invoice, the deadline has passed, and the client politely asks whether you are still interested. The meeting was never the problem. The step from notes to action is — and that is exactly the step AI handles well, if you build it as a routine rather than a party trick.
Notes capture moments, not commitments
A note says “VAT question”. A task says “call Mr Brunner by Thursday and confirm the VAT rate for the March offer”. Notes are compressed memory for the person who was in the room; tasks are commitments that survive a busy week. In a larger company, a project manager converts one into the other. When you are self-employed, nobody does — and that is where client work quietly dies. Not laziness: a missing step in the workflow.
The routine: ten minutes, three steps
- Dump everything, unfiltered. Right after the conversation — or as a twice-weekly pass — collect what actually exists: the scribbled page, the voice memo from the walk back to the office, the photo of the whiteboard, the half-sentence in a chat. Do not tidy up first; tidying is the machine’s job.
- One structured prompt. Ask for three separate lists: action items (task, owner, deadline if one was genuinely agreed), decisions made, and open questions. The crucial instruction: wherever the notes are vague, the AI must write “unclear” — not guess.
- Verify, then move. Walk the list once against your own memory of the conversation. Then put the tasks where tasks belong — your to-do list or calendar, not the chat window nobody reopens on Thursday morning.
Where AI genuinely helps — and where it invents
The strengths are real. It is relentlessly structured, it catches items you forgot you even wrote down, it never tires of formatting, and it drafts the two-line recap to the client (“Here is what we agreed; the offer follows by Friday”) that makes a one-person business look organised.
The weakness is just as real: language models fill gaps. If your note says “clarify VAT”, the model cheerfully writes “clarify VAT by Friday” — a deadline nobody agreed to. An action item with an invented date is worse than no action item, because you now trust it. Two rules keep the output honest: every deadline is confirmed against the source, and anything marked “unclear” goes back to the client as a question — not into your task list as a guess.
Client data stays out of the prompt
Meeting notes are among the most sensitive documents a self-employed person holds: names, figures, deadlines, sometimes health or legal details. The revised Swiss Data Protection Act (nDSG) does not forbid AI here — it obliges you to choose tools and providers deliberately. In practice: replace names with “Client A” and amounts with placeholders before pasting into a cloud tool, or run a local model that never leaves your computer. Turning notes into tasks does not need real names; “Client A, offer by Friday” works just as well.
“Here are my raw notes from a client meeting. Produce three lists: (1) action items — task, owner, deadline, only where a date was explicitly agreed; (2) decisions made; (3) open questions. Write ‘unclear’ wherever the notes are vague. Do not invent dates, names or amounts. Keep it short and factual.”
The habit that makes it work
Run the pass within a few hours of the meeting, while your memory can still check the output — a week later it is archaeology, not administration. Done this way, the offer goes out on Friday as promised, the VAT question is answered before the client asks again, and the “what exactly did we agree?” evening disappears from your week. That is the entire point: not beautiful minutes, but kept promises.
