There is so much talk about artificial intelligence that it is hard to tell what already works from what is still a promise. For an SME, the useful question is not what AI can do in general, but which specific task in your business it can do better or faster, with what data and under what control. This article reviews the uses that work well today in small and medium-sized businesses, how to get started and what to watch out for.
Where AI fits well and where it does not
Today's AI is particularly good at tasks that involve a lot of reading, sorting and writing, and that until now were done by a person because the data did not arrive in order. It performs worse where total accuracy is needed without review, or where the rules are already clear and an ordinary program follows them without making mistakes.
| Fits well | Fits less well |
|---|---|
| Reading documents in different formats (invoices, delivery notes, orders) | Calculations with fixed rules, such as settlements or payroll |
| Sorting emails, incidents or enquiries | Decisions that affect people without human review |
| Drafting replies, quotes or reports | Texts sent as they are without anyone reading them |
| Summarising long documents or conversations | Data sources that nobody keeps up to date |
| Answering customers' frequently asked questions with reviewed information | Legal, tax or technical advice without supervision |
Specific uses that already work
Reading invoices and documents
This is probably the most direct use. Supplier invoices arrive as PDFs, scanned paper or photos, each with its own format. AI can read the supplier, date, amounts, taxes and lines, and have them ready to be recorded in the accounting or invoicing software. The person goes from typing to reviewing.
Sorting email and incidents
A general mailbox receives orders, enquiries, complaints and advertising. AI can classify each message, extract the key data and route it to the right person.
Automating repetitive processes
When a process consists of reading something in one place and noting it down in another, AI combined with automation can do it on its own, leaving the exceptions to a person.
Help with writing
Drafts of quotes, product descriptions, replies to routine enquiries or meeting summaries. It saves time as long as someone reviews the text before it is sent.
Questions about your own data
Asking in plain language things like "which customers have reduced their purchases this quarter" based on the data in your management software. It works well when the data is well organised and the tool can only read what it should.
How to start without wasting time
- Choose a task, not a technology. Look for something repetitive, high-volume and currently done by a person reading and typing.
- Measure how it is done now. How long it takes and how many errors there are. Without that baseline you will not know whether things have improved.
- Test with real documents. Your own, with their flaws and odd formats, not clean examples.
- Keep human review. At first, someone checks everything. Over time, only what the system flags as doubtful.
- Connect the result. If the AI reads the invoice but the data then has to be copied by hand into the accounts, the saving is halved.
- Expand when it works. One case that works is more convincing than five half-finished pilots.
Data, privacy and regulation
Using AI does not change your obligations regarding data; it makes them more visible.
- Personal data. If the system processes customer or employee data, the General Data Protection Regulation (Regulation (EU) 2016/679) applies. Check where the data is processed, whether the provider uses it for other purposes and that there is a data processing agreement. The Agencia Española de Protección de Datos (the Spanish Data Protection Agency) publishes guidance on AI and data protection on its website.
- Free tools. Pasting a contract, a payslip or the customer list into a public chat is an easy way to lose control of that information. Define what can be used and for what.
- European AI Regulation. Regulation (EU) 2024/1689, on artificial intelligence, regulates these systems according to their risk. The uses in this article are not usually among the high-risk ones, but the regulation requires, among other things, that companies using AI take measures to train their staff in this area (Article 4) and that people are informed when they interact with an AI system, such as a customer service assistant (Article 50). The regulation was amended in 2026 and some deadlines have changed, so consult the current consolidated text or an adviser before using AI in sensitive processes, such as recruitment.
Common mistakes
- Starting with the most ambitious case instead of the most repetitive one.
- Trusting the result without checking it, especially amounts and dates.
- Setting up the AI as an island, with no connection to the programs where the data needs to end up.
- Not deciding who is responsible for reviewing errors and improving the system.
- Letting each employee use whatever tool they like with company data.
How we do it at Infonet
At Infonet Consultores we adopt each new technology, from the cloud to artificial intelligence, and in 2026 we apply it with Capturia, which reads your invoices and documents, records them in your invoicing or accounting software and automates repetitive processes between your systems. When the case is more specific, we solve it as a bespoke development linked to what you already use, and we also help with GDPR compliance. You can learn about the team's track record at software company in Huelva.
