Where to start with AI in a small company

When an owner hears “adopt AI”, they usually picture a big project: a review of the whole company, new systems, staff training and a budget for a year. So the conversation is often postponed, although you can start small.

Why not rebuild everything at once

A full rebuild looks good on paper, but you pay before you see any benefit. And over months of work the company changes, and what was the main thing at the start loses importance by the end.

People need time to get used to it. If you change everything at once, many will go back to the old way, and the system will sit idle.

What to do instead

We suggest starting with one task that has two properties: it clearly gets in the way of work, and it can be solved quickly.

Write down on a sheet of paper everything that irritates you and your team: manual reports, copying data from emails into a spreadsheet, the same client questions, preparing invoices and completion certificates. Next to each item put two ratings from one to ten: how much it hurts and how hard it is to change. Take the item where the pain is high and the difficulty is low. The first solution may be imperfect; the main thing is that people use it.

Which tasks usually fit

• Invoices and completion certificates: the model prepares a draft from the data, and a person checks and confirms it.
• Voice messages instead of manual entry: you dictate, and a task appears in the calendar or a row in a spreadsheet.
• Typical client questions in a messenger: the answer arrives within a minute, even when everyone has gone home.
• Site summaries: who arrived, what equipment, what was delivered.

In each case a person spends time on mechanical work.

An example

Imagine a small company where every week one employee manually collects data for invoices and completion certificates. The work is tedious, mistakes happen, and when the person is on vacation, everything stops. It hurts, and it can be solved quickly: AI prepares drafts from data you already have, and the employee reviews and confirms them. That is where we start.

The phrase “make the company digital” will not make such a list. With that wording it is unclear what to do on Monday.

Common mistakes at the start

• Taking on the hardest thing because it seems the most important. The first result will come late.
• Inventing a task from scratch. It is better to take what people already do by hand every day: there you see right away whether things got better.
• Waiting for a perfect result. The first version handles the main part of the job, and the rest is finished along the way, once you see how it lives in daily work.
• Not appointing an owner. Even a small project needs a person who will say, “It works” or “It doesn’t work.”

What to check before you start

First think about data. If your work involves sensitive information, such as correspondence with clients or contracts, decide in advance what can be given to external services and what cannot. For lawyers and many others this is a matter of professional confidentiality.

The task should have a specific person who will find work easier. Then it is immediately clear whether the solution works.

Agree in advance on how you will know it worked: the task takes less time, errors are gone, the client gets an answer faster.

What next

If it worked, you see from your own example how much time it takes and where you need help, and you decide what to do next yourself. If not, you spent little and know what did not work.