Made for Swiss self-employed & SMEs
Every morning, the same scene: a full inbox and the quiet feeling that the real work will have to wait. For notaries, fiduciaries, doctors and small offices, email is not a communication tool — it is a to-do list that other people get to write. AI will not empty your inbox by magic. But it can read, sort and draft for you, so that you spend your morning deciding instead of typing.
This article describes a calm, practical workflow for the daily inbox. It works for anyone who runs a small practice or office in Switzerland: the fiduciary handling twenty client dossiers, the notary coordinating appointments and documents, the doctor who writes the same kind of reply every afternoon. No software project, no big setup. Just a simple routine with a human check at every step.
The inbox is a to-do list in disguise
Most people treat email as reading. That is the mistake. Every message contains either a task, a question, a deadline, or noise. When you read your inbox the way you read a newspaper, you end the hour with nothing finished and a vague feeling of falling behind. The better habit is to treat each message as a small work item and decide, quickly, what to do with it.
What AI can do — and what it cannot
AI is good at reading and summarising. It can look at a long client email and give you three lines: what the person wants, what they are asking for, and what you need to answer. It can sort messages into groups — urgent, needs a reply, information only — and it can draft a reply in your tone. What it cannot do is decide for you. It does not know your client, your obligations, or what is genuinely urgent in your week. So the rule for the whole workflow is simple: the machine prepares, you decide.
Four steps to a calmer inbox
The workflow has four steps. You can do them in one block in the morning, or spread them over the day. The order matters less than the habit of doing each one.
Step 1: Summarise
Feed the day's messages to your AI tool and ask for a one-line summary of each, plus the action it implies. For a long client email this saves real reading time; for a short one it is almost instant. The prompt can be as simple as the box below.
Here is today's inbox. For each email, give me: (1) who wrote, (2) what they want, (3) what they expect from me, (4) whether it needs a reply today. Keep each answer to one line. Do not invent details. Emails: [paste the messages]
The instruction "do not invent details" is the important part. You want the tool to read, not to guess.
Step 2: Prioritise
Once the summaries are in front of you, you decide the order. The AI can suggest a grouping — urgent, today, this week, information only — but the grouping is yours to confirm. A client asking for a document before a deadline is urgent; a newsletter is not. You know your week better than any model, and the two minutes you spend ordering the list is time you get back for the rest of the day.
Step 3: Draft the reply
For the messages that need an answer, let the AI draft. Give it the context it needs — the client, the topic, the tone — and let it write a first version. Then you correct it. This is where the workflow pays for itself: drafting a short, polite, factual reply takes you two minutes instead of ten, and the result is more consistent because it always starts from your own instructions.
Write a short reply to [client name] about [topic]. Tone: friendly and factual, in calm Swiss Standard German. Confirm the facts, answer the question, and propose one clear next step. Do not invent dates, amounts or names — leave a placeholder [..] where you need me to check. Context: [brief note about the matter]
Review the draft in two passes, as with any AI text: first the facts, then the tone. Send it from your own mail program so the correspondence stays in your record.
Step 4: Follow up
Some emails do not need a reply today — they need a reply next week. Instead of leaving them in the inbox, move them to a small follow-up list, or let the tool remind you. A fiduciary who chases an outstanding document three days after the client promised it looks organised; one who forgets for a month looks absent. The difference is a list, not a talent.
Data protection: client mail is not neutral
Client emails contain names, matters, sometimes health or financial details. Once you paste them into an AI tool, the Swiss Data Protection Act (DSG) applies. In practice that means three habits:
- Use a business-grade tool, not a private consumer account.
- Check the small print: does the provider store your input or train on it?
- Keep client data out of anything you have not approved for your practice.
If you are unsure, anonymise first — replace names with placeholders, draft the structure, and add the real names afterwards.
A calm weekly rhythm
None of this needs to be dramatic. A workable rhythm for a small office looks like this: once a day, run the day's mail through the summariser and sort it. Reply to what is quick, schedule what needs thought, and file the rest. Once a week, review the follow-up list so nothing slips. The goal is not inbox zero as a trophy. The goal is that the inbox stops running your day.
What stays human
AI can read your mail, but it cannot be responsible for it. The client relationship, the judgement about what matters, and the decision to reply now or later stay with you. That is the whole point: the machine handles the reading and the typing, and you keep the part of the work that actually requires a professional.
Start with one morning. Take the day's messages, run them through the summariser, and see how much faster you reach the first real decision. That single routine already removes the most draining part of a small office — and it builds the habit that makes the rest of the workflow feel natural.
