What holds pen-and-paper businesses back from AI isn’t money

Plenty of businesses in Spain run on a notebook, a pad of delivery notes and one person who knows by heart what every customer ordered last month. That isn’t backwardness or carelessness. It’s a system that has worked for years, and nobody has had a compelling reason to change it.

When a business like this starts thinking about artificial intelligence, it’s easy to assume that price is what’s holding it back. The official data points elsewhere.

In one sentence: according to the Spanish National Statistics Institute (INE) survey, as compiled by the National Observatory of Technology and Society (ONTSI) with 2025 data, the main reason Spanish companies that considered using AI didn’t go ahead is a lack of expertise (82.4%). Cost comes fifth (47.4%).

What the data shows

These are the reasons given by companies with 10 or more employees that considered using AI at some point but never went ahead. Each company could give more than one.

Source: ONTSI, Indicadores de uso de inteligencia artificial en España, 2026 edition, based on INE’s survey on ICT usage and e-commerce in enterprises (2025 data).

Two caveats for reading the table properly.

First, it’s a small group. Among companies with 10 or more employees that don’t use AI, only 11.1% have ever considered it. Before any barrier on the list there’s a quieter one: most haven’t even got as far as asking the question.

Second, the order holds when the base changes. Eurostat, which calculates the reasons over all Spanish companies with 10 or more employees that don’t use AI, arrives at the same ranking: lack of expertise (9.1%), data protection (5.9%), legal uncertainty (5.7%), data quality (5.3%) and, last of all, cost (5.2%). Two different ways of measuring reach the same conclusion.

ONTSI also notes that lack of expertise is the top barrier for companies of every size, micro-enterprises included.

Why this weighs more heavily on a pen-and-paper business

There’s no official statistic on “businesses that work on paper”, so some caution is due. But a few indicators come fairly close.

• Size. AI is used by 57.5% of large companies, 30.2% of medium-sized ones, 17.4% of small ones and 12.8% of micro-enterprises (fewer than 10 employees).
• Sector. In construction (11.4%), food and beverage services (13.3%), the sale and repair of motor vehicles (14.2%) and transport and storage (14.9%), fewer than one in six companies with 10 or more employees uses AI.
• People. Only 1.7% of micro-enterprises employ an information and communications technology (ICT) specialist. For AI specialists, the figure drops to 0.1%.

That last point explains a lot. In most small businesses, nobody’s job is to understand these tools. The “lack of expertise” isn’t a lack of ability: it’s that whatever knowledge there is depends on the owner finding time to look for it between one order and the next.

The five barriers, one by one

1. “I don’t know where to start”

What’s behind it. It’s rarely a lack of intelligence or willingness. What’s missing is a concrete question. “Using AI” is far too broad to make a decision about on a Tuesday afternoon.

What does solve it. Making the question smaller. Not “How do I apply AI to my business?” but “Which piece of data do we copy out by hand twice every week?” A question like that can be answered, tested and measured.

What doesn’t. A generic course on tools before you know what you want them for. You’ll learn features you’ll have forgotten within a fortnight, because they’ll have nowhere to be applied.

2. “I don’t want to put my customers’ data just anywhere”

What’s behind it. A sensible fear. Someone who has looked after their customer relationships for years doesn’t want to hand that information over to strangers.

What does solve it. Distinguishing between using an AI tool in a personal capacity and using it as a business. When a provider processes personal data on your business’s behalf, the General Data Protection Regulation (Article 28) requires that processing to be governed by a contract or other legal act. In practice, that means using business accounts, actually reading the terms and, in the meantime, starting with tasks that need no personal data, such as tidying up a product catalogue or drafting a template quote.

What doesn’t. Banning it and looking the other way. More than half of internet users in Spain (55.2%) already use generative AI, the vast majority with free tools. It’s reasonable to assume someone on your team has already tried it. Better to decide how it’s used than to assume it isn’t.

3. “I’m not sure it’s legal”

What’s behind it. A lot of noise. There has been so much talk about the EU AI Act that any use seems to require a lawyer.

What does solve it. Knowing that the Act classifies uses by level of risk and that most of what a small business would do (drafting, summarising, extracting data from a delivery note) is not high-risk. For a small business that uses AI, two things matter most today:

• Training your team. Article 4 requires those who use AI systems to take measures to support AI literacy among the people operating them. The 2026 amendment (Regulation (EU) 2026/1744) softened the wording and made clear that no specific level has to be guaranteed for each person, but it did not remove the obligation.
• Transparency. Since 2 August 2026, AI systems that converse with people must inform them that they are talking to an AI, unless this is obvious (Article 50). That obligation falls mainly on whoever develops the system, so if you use a provider’s assistant, it’s worth checking that it complies.

Obligations for high-risk systems under Annex III, which covers, for example, using AI to select staff, have been postponed to 2 December 2027. If your idea goes in that direction, get advice first.

What doesn’t. Waiting for “everything to become clear”. The part that affects everyday use is already clear and already applies.

4. “Everything we have is on paper”

What’s behind it. The idea that you have to digitalise everything first and that AI comes afterwards, as a reward. And the feeling that this means months of typing paperwork into a computer.

What does solve it. Flipping the order. Today’s assistants can read a photo of a delivery note, an invoice or a handwritten note, and transcribe a voice message. Paper can be the input, not the obstacle. A photo taken on a mobile phone can become a tidy row in a spreadsheet without anyone typing it in.

The same goes for the incompatibility barrier (35.4%): there’s no need to replace what already works. Often, a small layer that picks up information from where it is and puts it where it’s needed is enough.

What doesn’t. Thinking it will read everything perfectly. Difficult handwriting, an abbreviation only its author understands or a badly taken photo will produce errors. That’s why, at first, someone has to check what the AI has read.

5. “It’s expensive”

What’s behind it. A real barrier: almost half of companies cite it. It isn’t an excuse.

What does solve it. Separating the price of the tool from the real cost. In a small business, what weighs most is usually time: learning, testing, correcting and supervising. That’s why it makes sense to start with a single task, with a scope that can be assessed within a few weeks, before committing further.

As for public funding, it helps to know where things stand. In Catalonia, ACCIÓ, the regional business competitiveness agency, has already closed its 2026 call for AI adoption vouchers, which funded an advisory service. At national level, the Spanish Government’s IA360 Plan includes a business AI voucher, with a pilot in the first half of 2027 and a general call before the end of that year. The plan itself states that it will not fund mere licence subscriptions. It may prove useful later on, but there’s no point waiting for it before taking the first step, because the first step costs nothing.

What doesn’t. Running the numbers without knowing which task you’re going to change. Without that, any quote will look either expensive or suspiciously cheap.

What AI won’t do in a pen-and-paper business

To complete the picture, it’s also worth saying what doesn’t change:

• It won’t organise a process nobody has defined. If everyone does the same task in a different way, AI will automate that mess.
• It won’t replace whoever knows the customers. It can take repetitive work off their hands; it can’t know what that person knows.
• It makes mistakes. It needs someone responsible for checking, especially at first.
• It isn’t worth it for everything. A task that comes up twice a month is probably not worth the effort of changing.

A first step you take with pen and paper

If the main barrier is not knowing where to start, the first step isn’t technological. It’s observation. And in a paper-based business, the most natural way to do it is on paper.

For one week, leave a sheet on the work desk and jot down every time one of these things happens:

1. Someone copies the same piece of data into two different places.
2. Someone spends more than five minutes looking for a piece of paper.
3. Someone asks about something that’s already written down elsewhere.
4. A document has to be redone because information was missing.

At the end of the week, look at what comes up most often. That’s your concrete question. With it, any conversation about artificial intelligence starts somewhere very different: in your operations, not in the tool.

Frequently asked questions

What is the main barrier to AI adoption for Spanish SMEs?

A lack of relevant expertise. According to 2025 INE data compiled by ONTSI, it is cited by 82.4% of companies with 10 or more employees that considered using AI and didn’t. It is also the top barrier for micro-enterprises.

Is artificial intelligence expensive for a small business?

Cost is a real barrier, but it ranks fifth (47.4%), behind lack of expertise, data protection, legal uncertainty and data quality. In a small business, the biggest cost is usually the time spent learning and supervising.

Do you need to digitalise your business before using AI?

Not necessarily. Current tools can read photos of documents and transcribe voice notes, so paper can be the starting point. Their output should be checked, because they make mistakes with difficult handwriting or poor-quality photos.

What does the EU AI Act require of a small business that uses AI?

For most everyday uses: taking measures to support AI literacy among the people who use it (Article 4) and making sure that, when a customer converses with an AI system, they know they aren’t talking to a person (Article 50, applicable since 2 August 2026). High-risk systems under Annex III carry additional obligations, postponed to 2 December 2027.

Sources

• ONTSI (Red.es). Indicadores de uso de inteligencia artificial en España. Edición 2026, datos 2025 (in Spanish).
• INE. Survey on ICT usage and e-commerce in enterprises (in Spanish).
• Regulation (EU) 2016/679, General Data Protection Regulation, official text, Article 28.
• Regulation (EU) 2024/1689, Artificial Intelligence Act, official text.
• Regulation (EU) 2026/1744, Digital Omnibus on AI. Official Journal of 24 July 2026.
• ACCIÓ. ACCIÓ innovation vouchers 2026 (in Spanish).
• La Moncloa. Plan IA360 (in Spanish), September 2026.

If you try the one-week exercise and want to talk through what comes out of it, you can write to rafael.vidal@isenzo-ia.com.