Artificial intelligence is everywhere in the news, and most business owners feel they “should be doing something”. The problem is that the conversation usually starts with tools: which model, which app, which subscription. For a small or medium-sized business the right starting point is different: the work that happens every day and costs time.
Start with the work, not the tool
Before opening any app, list the tasks that repeat every week. A task is a good candidate for AI when:
- it repeats often, daily or weekly,
- it takes real time from people who could be doing something more useful,
- it follows rules or patterns, even if they are not written down anywhere,
- it relies on text, documents or data you already have.
If a task meets three of the four, it is worth a closer look.
Where it usually pays off
In small and medium-sized businesses, the first results almost always come from the same areas:
- Customer questions: stock, delivery times, prices, instructions. An assistant can answer the routine ones and hand the rest to a person.
- Documents: invoices, orders, delivery notes, contracts. Reading them and preparing the entry happen automatically, while a person reviews and approves.
- Quotes and replies: first drafts of quotes, email replies or product descriptions, which staff correct instead of writing from scratch.
- Internal knowledge: procedures, manuals, technical sheets. An assistant that answers from your own documents saves the “where do I find this?” questions.
- Reports: figures from your systems, asked for in plain language.
Three low-risk steps
1. Map it. Pick the two or three tasks that cost the most time. Note how long they take today and how often they happen. That is your baseline.
2. Pilot one process. Test the solution on a single task, with real data, for a few weeks. Measure something specific: hours saved, errors, response time to customers.
3. Integrate. If the pilot works, connect the solution to the systems you already use: ERP, CRM, online store, email. Without that connection, AI stays one more window on the screen that nobody opens.
What to watch out for
- Personal data: don’t paste customer or employee details into free tools without knowing where they end up. The GDPR applies here too.
- Human oversight: anything that goes to a customer or into your books should be checked by a person, at least at first.
- Running costs: many solutions charge per use. Estimate the monthly cost at your real volume before you commit.
- Vendor lock-in: prefer solutions where your data and settings can move if you need to change tools.
- Regulation: the EU Artificial Intelligence Act sets obligations depending on the use. For most everyday applications they are limited, but it is worth knowing which category yours falls into.
What you can do tomorrow
- Ask each team to write down its three most repetitive tasks.
- Roughly estimate how many hours a week each one costs.
- Pick one for a pilot and decide in advance how you will judge whether it worked.
AI does not need to be a big project to pay off. It needs to solve a real problem, be measured, and fit into the way you already work. If you would like help with the mapping or the first pilot, see how we approach the move to AI.