Most conversations about AI in Swiss SMEs start with the same question: "Where do we even begin?" The honest answer is smaller than most people expect. You begin with one workflow — not a strategy deck, not a platform licence, and not a company-wide transformation programme.
Start with the work nobody defends
Look for the task that meets three conditions:
- It repeats weekly, in roughly the same shape.
- Nobody chose it as their career. Nobody will fight to keep it.
- A mistake is visible, but recoverable — an approval step catches it.
Classic examples: retyping enquiry details into a CRM. Drafting the same engagement letter. Answering "where is my order?" for the hundredth time. Chasing an approval sitting in someone's inbox. These tasks never appear on a strategy document, but they quietly consume whole working weeks.
A useful benchmark: fifteen minutes, twenty times a week, across forty-eight working weeks is 240 hours a year — six working weeks of capacity spent on one repetitive task. That is the scale hiding inside workflows your team considers "just how it works".
Pick the workflow where the numbers already exist
The first automation should pay for itself in a way your accountant can see. That means:
- Measurable before and after. Response time, hours per week, error rate. If you cannot measure it today, you will not be able to prove the gain tomorrow.
- A clear human checkpoint. The automation drafts, routes, or classifies — a person approves. This keeps quality under control and builds trust with the team.
- Boundaries agreed up front. Which systems it touches, which data it may read, where the output goes, and who is accountable when something looks wrong.
This last point matters more in Switzerland than almost anywhere else. The FADP is not an obstacle to AI automation — but it does mean data flows should be deliberate: which customer data enters which system, where it is stored, and who can access it. A well-scoped first project answers these questions before a single line of integration code exists. Swiss data hosting is available where local residency matters.
What the first project usually looks like
In practice, a first automation at a 10–50 person firm is often one of three patterns:
Enquiry intake. A form, inbox, or chat message becomes a structured CRM record, a drafted reply, and a routed task — with a person checking before anything is sent.
Internal answers. An assistant trained on approved business knowledge — policies, price lists, product specs — answers routine questions from staff and customers, and escalates what it is not sure about.
Document drafting. Quotes, offers, or standard letters assembled automatically from structured inputs, reviewed and released by a human.
None of these requires new infrastructure. They run on the tools you already use: your email, your CRM, your document store.
The order of operations
- Audit. Thirty minutes. Map the repetitive work, the systems, and the data constraints. Rank the opportunities by value and feasibility.
- Pilot. One workflow, fixed scope, a few weeks. Prove the numbers on something small enough to finish.
- Scale. Only after the pilot delivers. Automate the second workflow with the governance and patterns the first one established.
The sequence matters more than the technology. A SME that pilots first builds internal trust, keeps risk contained, and ends up with automations the team actually uses. A SME that starts with a platform licence ends up with a platform licence.
The bottom line
The right first automation is rarely the most impressive one. It is the one that removes a task nobody will miss, produces numbers your team believes, and establishes a pattern you can repeat. Find that workflow, scope it tightly, keep a human in the loop — and let the results argue for the second project.
Curious which of your workflows qualifies? The free AI Readiness Audit maps your recurring tasks and ranks what is worth automating first.