Generative AI in the company: concrete use cases

In two years, generative AI has gone from a technological curiosity to an everyday work tool. But between the enthusiastic headlines and the reality of an SME there is a practical question: where does it create real value and where does it just risk wasting time? The answer is not “everywhere”. It is in specific areas, where producing text, summaries or drafts is a bottleneck. Let’s look at the concrete use cases, the limits and the conditions for using it without getting hurt.

What generative AI actually does

Generative AI produces new content — texts, summaries, drafts, code, images — starting from an instruction. It does not “know” anything in the sense that a professional knows it: it recognises linguistic patterns and generates the most probable output. This has two practical consequences. It is excellent at producing a first version of something. It is unreliable as a source of truth, because it can invent facts, figures and references with perfect confidence.

The operational rule that follows is simple: use it to speed up production, never to establish what is true. Substantive judgement remains human.

Concrete use cases in an SME

There are areas in which generative AI significantly reduces repetitive work.

Communication and marketing: drafts of newsletters, product descriptions, posts, website copy. You start from a generated draft and refine it, instead of from a blank page.
Document summarisation: summaries of minutes, long contracts, reports. The AI extracts the key points; the person verifies the ones that matter.
Assisted drafting: first versions of offers, sales e-mails, standard customer service replies, specifications.
Translation and revision: a first translation between Italian, German, French and English — useful in a quadrilingual country — followed by human proofreading.
Internal support: answers to frequently asked questions from staff, starting from company documentation.
Qualitative analysis: classification of customer feedback, identification of recurring themes in large volumes of text.

In all these cases the benefit is not “AI does the work”. It is that it lowers the cost of the first draft, and the first draft is often the slowest part.

Where NOT to use it without oversight

There are areas in which relying on raw output is dangerous.

Tax and legal advice: AI can cite non-existent or outdated rules. In Switzerland the rules on VAT, taxes and company law are specific; a plausible but wrong answer is costly.
Precise data and calculations: it is not a reliable calculator. Figures, rates and deadlines must be verified at the source.
Decisions: AI proposes, it does not decide. Responsibility for a business choice cannot be delegated to a model.

The critical issue: data protection

This is where many SMEs go wrong. Pasting a contract, a client list or personnel data into a public AI tool potentially means transferring data to a foreign provider, often outside Switzerland. The new Swiss Data Protection Act (nFADP) requires care in the processing and communication of personal data, and for client and staff data this is not a detail.

The minimum rules to set for yourself:

Do not enter personal or confidential data into free public tools.
Use enterprise solutions that guarantee that the data is not used for training and that clarify where it is stored.
Define an internal policy on what can and cannot be fed to AI.

How to set up adoption

Adopting generative AI does not mean buying a tool and hoping. It means an orderly path.

Start from a problem, not from the technology: identify a real bottleneck (e.g. offers take too long) and try AI there.
Measure: how much time do you really save, net of revision? If proofreading cancels out the gain, the use case is not a good one.
Train people: the value depends on how instructions are written and on how good you are at recognising a wrong output.
Keep human oversight over everything that leaves the company or affects numbers and obligations.

How we work

SIH ONE is an AI-native fiduciary group: we use generative AI in our processes to speed up summaries, drafts and analyses, but every output that touches numbers, obligations or decisions is validated by a licensed professional who retains responsibility for it. We help you identify the use cases that make sense in your reality and set them up in a way that complies with Swiss data protection.

Want to understand where generative AI can really save you time? Let’s talk about it together.