AI for a small office, without the big project

The real problem

Small teams are already busy. The week fills up with enquiries, appointments, quotes, supplier mail, and the quiet work of keeping client details straight. A new tool has to earn a place in that week. If it does not, it becomes one more thing to check.

Tools also pile up. Someone tries a chat window for a draft. Someone else starts a second subscription because the first one missed a job. A folder of half-finished prompts sits next to the inbox. None of this is a failure. It is what happens when a useful idea arrives before anyone has agreed what job it is for.

An "AI project" can sound expensive and risky on top of that. A long rollout, a new platform, a promise that the whole business will work differently. Most small offices do not need that, and they should not be talked into it. They need a few repetitive jobs to take less effort, and a way of working the people already there can keep up.

Where we stand

Building has been a large part of our own work with AI. Even so, building is only a small part of the whole picture. Choosing the right tools for the real needs, setting them up well, getting people to use them, and keeping the work organised and consistent matter far more.

Start from the work, not the tool

Before anyone compares products, write down the jobs that already repeat. Three to five is enough. If a task happens every week and a careful person would answer it in much the same way, it is a candidate. If it needs a fresh judgement every time, leave it with a person.

• Enquiry replies that should sound like the business, not like a different person each time
• Scheduling, and the short messages that go with a change of time
• Follow-ups that get dropped when the week gets busy
• Meeting notes that should land in the client file, not stay in a notebook
• Quotes and other documents that should follow the same shape
• Client details that should match, whether they live in an inbox, a spreadsheet, or a shared folder

You do not hand all of these to an assistant. Pick one. Write down what a good result looks like, including what must never be sent. Only then ask a tool to help. The tool is there to support the people who already know the clients. It is not there to replace that knowledge.

What an always-on assistant can do

A newer kind of assistant stays available, instead of waiting for someone to open a chat and start from a blank box. In a small office that can mean a steadier rhythm: the same checks, the same tone, and a reminder that a follow-up is still open.

The three below are current examples of that idea. They are not a shopping list, and we are not suggesting you take all of them on. Read what they actually do, then come back to the job you have.

Product details as of October 2026; these tools change quickly.

What this can look like

These are illustrations, not client case studies. They are not a claim about a named firm. They show the shape of the work in two ordinary offices.

Ground rules

A few checks keep this workable for businesses in Germany, Austria, Switzerland, and the Nordic countries.

This is practical guidance from our work, not legal advice.

For clients in Switzerland, data protection follows Swiss law (revFADP). Check where the data is processed, and whether a data processing agreement is available, before client or staff details go into a tool.

• Check that the product is available in your country. Muse, for example, is US only today.
• Do not let an assistant send anything to a client without approval, at least at first. Approval can stay in place for anything that leaves the office.
• Start with one assistant and one workflow. Add a second only when the first one is actually in use.

Give the team short usage rules, and keep a short record of them. Our AI training is where we write those rules down with the team.

How we help

We do not start with a platform. A short AI briefing is a few hours on what these tools can do for this business, which risks are worth naming, and the next steps. A setup sprint is for when the job is already clear: one or two tools, or one workflow, set up for a real task, with a check-in once it has been in use. A team workshop sits between them when several people need the same rules.

The ranges below are the same examples as on our AI training page. They are there so you can plan. The price you accept is a fixed quote, agreed after a free fit call. If we are not the right team, we will say so.