Short answer
In a one-person business, AI is not an employee. It is a first-draft machine and a sparring partner — and it is good at exactly the tasks that would otherwise never get done, because they are tedious, not because they are hard.
It handles drafts, structure, summaries, proofreading, research groundwork and code. It cannot handle judgement, responsibility, the relationship with your customer or knowing what is true. That part is still yours, and it becomes more important, not less.
What it actually does
I run a sole trader business with talks, coaching, a publishing house and a number of my own systems. All the administration lands on the same person. This is where AI has helped most — measured in things that actually get done, not in hours saved on paper:
A first draft of everything. The quote, the email to the head teacher, the product text, the post. A draft is not finished — but the difference between a blank document and a half-decent draft is the difference between “tomorrow” and “now”.
Structure out of the mess. Paste in the notes from three meetings and ask for a coherent summary with the open questions listed. This is the most underrated use, and it requires the least expertise.
Being the one who challenges. “Here is my plan — what have I missed, and what argues against it?” In a one-person business there is no colleague who tells you the idea is bad. It is no replacement for a real colleague, but it is a great deal better than nothing at all.
Technical work you would otherwise have bought in. This site has been built, reviewed and maintained with AI as a tool — including accessibility checks and structured data that would otherwise have cost consultant hours.
The tedious and repetitive. Renaming files, compiling material, proofreading, translation. Low risk, immediate benefit, and nobody misses the work.
What it cannot do
Knowing what is true. A language model produces what is probable, not what is verified. It invents sources, figures and paragraph numbers with the same confident tone it uses when it is right. Everything that goes out to a customer must be checked by you — and things you cannot check you should not publish.
Responsibility. If your name is on it, it is yours, whatever wrote the first draft. That goes for the quote, the invoice, the advice to the client and the text on the website.
Your voice. Default output from a language model reads like a brochure, and customers recognise it within three lines. Either you rewrite it into your own language or you don’t use it. A text that sounds like everyone else’s sells worse than no text at all.
The relationship. Nobody books a speaker because the email was well written. They book because they believe the person can deliver. That part cannot be automated, and the more everyone else automates it, the more valuable it becomes.
How I use it myself
I have built a small team of AI colleagues, each with a clear area — one who writes, one who reviews code, one who keeps track of the numbers, one who watches the outside world. It sounds more advanced than it is. The point is not the technology, but that each role has a bounded assignment and a description of how it should work.
That is also the whole secret of why it works. A general question gets a general answer. A role with a clear brief, clear limits and access to the right material produces something useful. Here is the team and how it is built, with three principles that govern everything: one thing at a time, the human decides, and always my voice.
Three common mistakes
1. Buying tools before having problems. The most common route in is “everyone is talking about AI, we should be doing something”. It ends with three subscriptions and no change at all. Start from the other end: which task takes time every week, is tedious, and carries little risk if it comes out half-decent? That is where your first automation lies.
2. Pasting customer data into anything. Personal data in a service where you don’t know where it is stored is a GDPR problem, not a detail. If you have no control over where the data ends up, it should not go there — case notes, client details and personal identity numbers least of all. I have written more about that in AI and GDPR.
3. Publishing unseen. It is easy to produce a lot, quickly. That is why the quality check — yours — is the only thing that separates a firm that has become faster from one that has become worse.
What it takes to get started
Less than most people think. A modern language model covers most of what is described above, and the big hurdle is not the technology but formulating what you actually want done.
If you need help with exactly that: an AI strategy starts with me at SEK 15,000 and answers what is worth doing in your business and in what order. A half-day course costs SEK 8,000, and policy, GDPR and EU AI Act work costs SEK 12,000. The first call is free of charge and exists to decide whether there is anything to do at all. Sometimes the answer is no, and that is a perfectly good answer.
The most important thing
If you only take one thing away: AI does not turn a one-person business into a company with ten employees. It makes you faster on drafts and removes some of the tedious work.
That is less than the promises in the adverts and far more than nothing — and for someone who is alone with everything, that difference is often what decides what gets done in a week.
Best wishes, Mikael GlännströmWellbeing coach, author and father of two