Are AI agents ready for Swiss SMEs?

Short answer: yes, for well-defined work. Not yet for running a business on its own.

An AI agent is more than a chatbot. It reads, decides on a next step and acts: it sorts an inbox, pulls the figures out of an invoice, drafts a reply, prepares a booking. In 2026 this moved from experiments into daily use. What has not changed is that agents are only as good as the process you give them.

Where agents work well today

• Documents in, structured data out. Invoices, receipts, tax documents, contracts: reading them and filling a system is repetitive, rule-based and easy to check.
• First drafts. Replies to common customer questions, offers from a template, meeting notes. A person reviews and sends.
• Preparation, not decisions. Collecting what is missing before a case starts: which documents are there, which are not, what to ask the client.

Where they still need a human

Anything that costs money, goes to a client or cannot be undone. Agents make mistakes that look convincing. The answer is not to avoid them but to put a human approval at exactly these points.

How to start

1. Pick one process. Repetitive, frequent, with a clear result. Not "use AI", but "every incoming invoice is captured within an hour".
2. Write down the limits. What may the agent read? What may it change? What must it never do?
3. Keep the approval step. The agent prepares, a person confirms. Remove the step only when the error rate has been low for months.
4. Measure before and after. Minutes per case and errors per month. Without numbers, every project feels like a success.

Data protection in Switzerland

Since 1 September 2023 the revised Federal Act on Data Protection (FADP) applies. For an agent that touches client data, three questions matter: where is the data processed, is it used to train someone else's model, and can you explain to a client what happens with their information? Choose providers and settings so that you can answer all three in one sentence each.

What I take from it

The companies that benefit are not the ones with the most AI tools. They are the ones that know their processes well enough to hand a part of them over. That is the work we do at Skopa: start small, measure, then grow.