Guide · 7 min read

AI automation for SMEs of 10 to 50 people: where to start in Switzerland

Katia Ryze, an AI-generated character. Article prepared with AI and reviewed by the Ryze Systems team·6 June 2026

The owner of an SME on the Vaud Riviera called the Ryze team last year with a simple request: “I want to automate my company.” Twenty-two employees, a full order book, and the conviction that he had to “do AI” before the competition did. We spent the first hour talking about no tool at all. Just listing, process by process, where his time went.

The conclusion surprised him. Eighty per cent of his pain came from one single place: entering site job sheets, copied out by hand in the evening from photos taken in the field. The rest (the CRM, marketing, dashboards) was comfort. Useful, but not urgent. We automated that one process. In three weeks. The return on investment was visible in the first month.

The mistake most agencies sell

“Complete digital transformation.” That is the phrase that comes up in almost every proposal I am shown. A six-month project, a five-figure budget, a plan that touches sales, operations, support and reporting at the same time. On paper, it is appealing. In the reality of an SME of twenty people, it is a promise that cannot be kept.

The reason is simple. An SME has no project team. The owner steering the “transformation” is also the one who answers the phone, signs the quotes and puts out the fires on Thursday afternoon. Asking him to absorb change on four fronts at once guarantees that none will be adopted for good. The most powerful tool in the world is worth nothing if the team has gone back to WhatsApp after three weeks.

Automating a bad process does not improve it. It makes it fail faster, and on a larger scale.

The only metric that decides where to start

Forget what is fashionable. The right first project can be calculated. Take each repetitive task of your week and note two numbers: the hours it consumes, and how often it goes wrong (an oversight, a double entry, information that cannot be found). Multiply. The task with the highest score is your starting point. Not the most visible one. Not the most “AI-compatible” one. The most costly one.

For most of the SMEs in French-speaking Switzerland that we support, this process falls into one of three boxes: capturing requests that arrive through five channels at once, tracking jobs or sites in the field, or following up on quotes and unpaid invoices. Rarely marketing. Almost never “a chatbot”. If your pain is mainly finding customer information (who ordered what, which quote is awaiting a reply), the real problem is not AI. It is that Excel has reached its limit. We put a figure on this hidden cost in the real cost of Excel for an SME.

How to recognise a first project that only looks good

Some candidates shine and are nevertheless bad starting points. The “chatbot on the website” is the perfect example: visible, flattering, easy to show in a meeting. But if it handles three requests a week, it gains you nothing. Conversely, the ideal task is often invisible and boring: an employee copying figures from one screen to another, an inbox sorted by hand every morning.

Three signals give away the right first project. It comes back every day or almost. It follows a clear rule, one you could explain to a new employee in two minutes. And when it goes off the rails, it shows: a delay, an unhappy customer, a duplicate invoice. If a task ticks these three boxes, you have your starting point. If it ticks only one, move on to the next.

Take another case, very different from the Riviera site: a brokerage firm in Lausanne that received its quote requests by email, WhatsApp and web form, without ever centralising them. On a Monday morning, nobody knew how many requests were awaiting a reply. The first project was not “AI”. It was to bring those three channels into one place, with automatic qualification. The rest followed once that foundation was laid.

Clarify before you automate

Here is the part nobody likes to hear. If a process is vague, automating it freezes it in that vagueness. Before putting a machine on it, you have to know exactly who does what, in which order, with which decision rule. That is often where you discover that “the process” does not really exist: there are three ways of doing it, one per person.

That is good news. Half of the gain comes simply from this clarification. Automation then seals the clean version. It is also why a serious audit starts by mapping reality, not by running a tool demo.

What gets automated first, in real life

Back to the owner on the Riviera. His bottleneck was entering the site job sheets. The answer was not “a big system”. The Operations module of Ryze OS captures the photo taken in the field, the AI module extracts the information (customer, service, quantities) and files it in the right folder. The technician photographs his sheet as before. The evening copying, however, has disappeared.

The decisive point: we touched only that. The CRM came later, once the team trusted the first piece. This is what we call a sprint approach: a tight scope, delivered in a few weeks, that proves its value before anything is widened.

Most of the business owners the Ryze team meets have heard of AI, automation and digital transformation. But when we sit down and start talking about specifications, things quickly become guesswork. Not because they are not capable, but because they do not really know what is possible with AI applied to their trade. For them, it is still something abstract, something they read in an article or saw in a video, but whose concrete limits they have never touched in their own context. So what we have learned to do is to forget the specification in the classic sense. We do not start from a thirty-page document. We start from an audit, a conversation, questions about what takes time, what is irritating, what regularly goes wrong. And it is during this audit that the spec is built, not before. As we identify a process, the owner understands what AI could do with it, and that unlocks three others he had not thought of. What we have experienced first-hand is this: two or three weeks after a first session, the owner calls back. "Ah, I hadn't thought of that one, could we do that as well?" Two months later, a new message, a new idea, a new process he has just seen differently. That is normal and healthy, because the understanding of what is possible comes gradually, not all at once. That is also our role: to open up the field of possibilities during the audit, so that the owner can identify what his company really needs.

The budget, in plain words

A useful first automation sprint, for an SME of this size, is around CHF 4’500 as a one-off, plus a monthly subscription for hosting and maintenance. It is neither five figures nor six months. It is a precise scope, delivered in two to four weeks. What makes this amount go up or down is detailed in custom software cost in Switzerland.

And the subscription is not billed per user. Whether five people or forty use it, the price does not move, because what you pay for is the infrastructure and the AI usage, not seats. This distinction weighs heavily in the long run, and it is the whole subject of the difference between an in-house infrastructure and an AI agency.

Why the sprint beats the transformation

A sprint has a virtue that big projects never have: it produces proof. After three weeks, your team is no longer discussing a promise on a slide. It is using something that works, and it knows it. This proof changes everything for what follows, because it creates buy-in. The second module is deployed in a team that has already seen the first one hold, not in a wary team.

The all-at-once transformation, by contrast, asks you to take things on trust for six months before you see anything. If a single assumption was wrong at the start (and there always is one), you find out at the end, when it is expensive to fix. The sprint fails small, early and without serious consequences. You adjust and carry on. It is less spectacular than a grand plan. Above all, it is much safer.

Your next step fits in one sentence

Do not choose a tool this week. Sit down for thirty minutes with a sheet of paper and list your repetitive tasks, with the hours lost and the failures for each one. The winner of that list is your first project. The rest waits its turn, and it will wait perfectly well.

If you want an outside view of that list, that is precisely what our audit does: thirty minutes to spot the high-ROI process and a proposal calibrated to your size. See what we have already delivered gives a concrete idea of the result.