AI for SMEs: where to actually start (no jargon)

The right question isn’t “how do I put AI into my business”, but “which concrete problem is costing me time or customers every week”. For a local SME, getting started with AI comes down to one simple rule: begin with a real case that is high impact and low risk, measure it, then scale it up. This is the approach we apply in our AI solutions.

Many business owners feel they are falling behind on AI, or overwhelmed by the surrounding noise. The reality is more reassuring: there is no technological “big bang” to orchestrate. The SMEs that succeed don’t roll out an abstract “AI strategy” — they fix one specific problem, draw a visible gain from it, then start again. Here’s how to frame that first step.

Forget the “AI strategy”, start with a problem

An AI initiative that starts with the tool (“we need a chatbot”, “let’s try this AI”) often ends up as a gimmick. Conversely, starting from a concrete pain point guarantees a useful project. Ask yourself three questions: which repetitive task takes us hours every week? Which questions do we answer over and over? Where are we losing requests for lack of responsiveness? The answer points to your first project — and makes it measurable.

To broaden the field of possibilities before choosing, browse our 10 concrete AI use cases for SMEs: they give an overview of the areas where AI is genuinely profitable for local businesses.

The most accessible entry points

Three families of use cases offer the best effort-to-benefit ratio for a first project, because they build on tasks already well identified within the company:

• Customer service — an assistant that answers frequently asked questions and qualifies requests, day and night, drawing on your real information.
• Automation — follow-ups, data entry, email sorting, syncing between software: time-consuming tasks delegated to an automated flow.
• Putting your documents to work — turning your procedures, sheets and records into a searchable knowledge base, so you can find the right information in seconds.

What these three entry points have in common: they relieve existing work rather than create new work. That’s what makes them easy for teams to adopt.

The golden rule: high impact, low risk

For a first project, weigh two criteria: impact (the time or revenue at stake) and risk (the consequences of an error). The right starting point sits in the “high impact, low risk” quadrant: a frequent, costly task whose imperfect answer carries no serious consequence. That way you avoid staking your first attempt on a critical process, while still securing a visible gain.

In practice, it’s better to automate the sorting of incoming requests — easy to correct — than to hand the AI a regulated decision or a sensitive accounting calculation from the outset. Trust is built case by case.

Measure, then scale

A first project is only worth it if you can tell whether it worked. Before you start, set a simple, observable indicator: hours saved per week, customer response time, number of requests handled without human intervention. There’s no need to chase the perfect metric — a qualitative benchmark is enough to decide whether to scale, adjust or drop it.

Once the first case is validated and its gain confirmed, scaling becomes natural: you reinvest the freed-up time into the next project. It’s this step-by-step logic, rather than one big project at once, that de-risks the investment.

Do you need a big budget to start?

No. The whole point of a first low-risk case is precisely to stay modest in cost and timeline. The budget then depends on scope — a simple assistant is nothing like a platform connected to several pieces of software. To gauge the orders of magnitude of an AI project in Switzerland, read our article on how much an AI project costs for an SME.

The costliest mistake isn’t picking the wrong first tool: it’s waiting for the “perfect” project while repetitive tasks keep eating up time. Starting small is already moving forward.