Made for Swiss self-employed & SMEs
A fiduciary in Lausanne keeps her client list in three places: an old address book, a spreadsheet she started two years ago, and the email signature of every person she has ever written to. When a client moves, changes bank details or marries, the information updates itself in only one of those places — and she never knows which one.
Every notary, fiduciary, doctor and small office runs on the same quiet infrastructure: a list of clients and contacts. It is rarely glamorous, and it is almost always a little bit wrong. Duplicates appear, details go out of date, and the same person exists under three slightly different spellings.
Cleaning it up by hand feels endless. But this is exactly the kind of repetitive, detail-heavy work a calm AI routine handles well — not to change anything automatically, but to prepare every suggestion so that you simply approve or correct.
Why the client list drifts into disorder
The mess is rarely anyone's fault. It is the natural result of how a practice grows:
- Duplicates: the same client is entered as "Müller GmbH", "Mueller GmbH" and "Müller AG", so three records share one history.
- Outdated details: addresses, phone numbers and bank details change quietly, and the old version keeps being used for months.
- Missing pieces: some contacts lack a postal code, others lack an email, and a few lack both.
The cost shows up in small, awkward ways: a letter returns as undeliverable, an invoice goes to the wrong address, or a client receives two reminders because two records were billed separately.
A calm routine in three passes
You don't need to clean the whole list in one heroic weekend. Three quiet passes, done once and then kept light, are enough.
Pass 1 — Merge the duplicates
Export your contacts as a plain list and let the AI group the entries that probably belong to the same person or company. It drafts the merge; you decide.
Here is an exported list of my clients and contacts. Group the entries that are clearly the same person or company — for example different spellings, old and new addresses, or a person and their company. For each group, show which record should be kept and which should be merged into it. Do not merge anything you are unsure about; put those in a separate list and explain why.
Pass 2 — Fill in the gaps
Next, the AI flags the records with missing or clearly outdated details and drafts what should be added — a missing postal code, a missing language preference, an old phone number.
Go through my contact list and flag every entry with missing or obviously outdated information: missing postal codes, missing email addresses, old phone numbers, incomplete company names. For each one, write what seems to be missing and, where the data allows, suggest the likely correct value. Never invent details — mark anything you cannot verify as "to confirm".
Pass 3 — Standardise the format
Finally, the AI brings the list into one consistent format: company names written the same way, addresses in the Swiss format, salutations and languages noted. This is the pass that makes future work easier.
What AI does well — and what stays with you
- Spotting duplicates and near-duplicates: finding the "Müller/Mueller" pairs that a human eye skips over.
- Drafting the corrections: proposing the right address, the right name, the right format.
- Keeping the work small: a tidy list is easier to keep tidy, so each future pass gets shorter.
What stays with you is every decision. The AI never writes into your client database on its own; it prepares, you approve. That is the whole point of a calm routine.
A note on data and security
Your client list is sensitive — names, addresses, dates of birth, sometimes health or financial details. It has no place in a public consumer chatbot. Use an approved setup: a local model or a zero-retention enterprise service, or anonymise the data first and add the real details back yourself at the end.
The outcome
One quiet afternoon is usually enough to turn a drifted, duplicate-laden list into a clean, trustworthy address book. The next invoice goes to the right address, the next letter reaches the right person, and you stop discovering — too late — that two records were really one client all along.
