The question I get most often isn't "Do we need artificial intelligence?" It's: getting started with artificial intelligence from where exactly, without wasting three weeks.
The answer lies in three characteristics that the first task must have. If one is missing, the attempt almost always fails.
The three characteristics
- You do it at least once a day. A weekly assignment doesn't generate enough repetition to see if it works. And most importantly, it doesn't force you to learn it.
- Can you tell me how much time it costs you?. Not "a lot": twenty minutes, an hour. If you can't quantify it, you'll never know if you've gained anything.
- One mistake is not serious. Your first assignment shouldn't be one where a mistake costs you a customer. You need a place where you can learn to recognize tool errors without paying for them.
The three together exclude, for example, customized quotes (a costly mistake) and include product descriptions, answers to frequently asked questions, call transcripts, and the first draft of posts.
The task that wins in most cases
For those who sell products, the first task is almost always the first draft of the cards, and it's a textbook case: repetitive, measurable, and with a safety net, namely proofreading before publication.
It also has a secondary effect that is worth more than the time saved: it forces you to line up the actual data for each reference, that is, to build that table of measurements, materials, and limits that was needed anyway.
Second place goes to repetitive responses to customers, for the same reason: the material already exists, it just needs to be organized.
How to measure without deceiving yourself
The honest measurement is done in three steps, and takes two weeks.
- Week zeroDo the assignment as usual, but time it. Five times is enough for an average.
- Week one and two: do it with the tool, also timing the correction time. This is where the truth lies: many savings disappear during the revision.
- In the endCompare the two numbers and also look at quality. If the time has decreased but returns have increased, you haven't saved anything.
A real gain of the 30% on a daily task is worth more than a theoretical gain of the 90% on something you do once a month.
What the numbers say about others
Among European small businesses, the share that uses AI is 17%, against the 55% of the great ones. And the most widespread uses are simpler than the word suggests: analysis of written language (11.8%), generation of images, video or audio (9.5%), generation of text or speech (8.8%), transcription (7.2%).
None of these require a project. They require an afternoon of setup and the discipline to proofread.
The gap between those inside and those outside today isn't technological. It's that someone has chosen a task and measured it, and someone else is still deciding which tool to buy.
If you want to think about what your first task is, write to me: this is the part where you waste the least time if you do it together.
Sources
- Eurostat, 20% of EU enterprises use AI technologies, announcement of December 11, 2025: adoption by size class
- Eurostat, Use of artificial intelligence in enterprises, Statistics Explained: Most Popular Technologies in Business
- Regulation (EU) 2024/1689 (AI Act), Article 4: Staff literacy
The Eurostat data concern companies with at least ten employees.