How it works today
- Scroll through the inbox each morning, sorting in your head — 25 min
- Answer the same question for the third time this week — 15 min
- Discover yesterday's truly important enquiry only in the afternoon
- Check the inbox again at night, because the feeling remains you missed something
How it works afterwards
- Inbox is pre-sorted: urgent, standard, can wait
- A draft is ready for the recurring questions — 2 min instead of 15
- Whatever truly needs you sits at the top and doesn't get missed
What we actually build
1. Understand your inbox
We look at a typical week and find the patterns: Which enquiries keep coming? Which are urgent even though they don't sound urgent? What does "important" mean for you — because that's different for everyone and can't be derived from a generic rule.
2. Define sorting rules
The patterns become categories with clear criteria. Not "important/unimportant", but concrete: existing client with an open order, new enquiry with a budget, supplier invoice, internal note, newsletter. Each category gets a treatment.
3. Draft replies for the repeats
For the five to ten questions you get constantly, drafts are built from your previous answers. Not as a rigid template — as a draft that picks up the specific enquiry and still sounds like you.
4. Approval stays with you
Nothing goes out automatically. That's not a technical limitation but a deliberate decision: one wrongly sent reply costs more than it ever saved. You read, you send.
Tools — and where it doesn't help
What gets used depends on your inbox. Microsoft 365 and Google Workspace offer solid built-in options that keep data inside your environment — the simplest variant from a data-protection standpoint. Where that's not enough, a separate model with clearly defined data flow is added.
Where it doesn't help: with enquiries that need real decisions — discounts, complaints, delicate client situations. The system sorts those to the top for you and writes nothing. That's exactly the point.
Email is the most sensitive data pot in the company
Your inbox holds client data, pricing information, sometimes health or HR data. The revised Swiss data protection act treats these categories with different degrees of strictness — and a tool that reads everything indiscriminately is not defensible in most cases. That's why we define, before implementation, which mailboxes and categories are included at all.
→ How we bring data protection and AI together
What this means in time
If you spend 40 minutes a day sorting and answering repeats, you conservatively gain 25 minutes per day — about two hours a week, roughly 95 hours a year. The bigger effect is usually a different one: the important enquiry gets answered in the morning instead of the afternoon.
Frequently asked questions
Does the AI reply to my clients on its own?
No, and that's built in deliberately. The system sorts and drafts. Sending stays with you. Fully automatic replies only become reliable once a process has run stably for months — and even then only for narrowly defined cases.
Does the AI read my entire inbox?
Only what we approve. In practice we narrow it to specific mailboxes or folders — for instance the general info address rather than your personal one. That reduces the data-protection question considerably while still covering most of the workload.
What if the sorting gets it wrong?
At the start it sometimes does — that's part of it. That's why the first two weeks run in parallel: the system sorts, you check, we correct the rules. After about two weeks it's reliably accurate.
Does this work in French and English too?
Yes. The models are multilingual, and with multilingual inboxes the benefit is even bigger — because sorting across language boundaries is exactly the work that costs the most headspace.
Show me a typical week from your inbox
In the free intro call we find the three patterns that cost you the most time. No sales pitch.