Artificial intelligence or automation: which to use and when

> Declaration of interest: BAI Business offers automations and AI assistants and publishes this article. It is signed by Héctor Redondo, CEO and founder of BAI Business.

Today I want to talk about a confusion I see all the time: calling anything that happens by itself "artificial intelligence", and not knowing when one or the other is the right choice.

In short: automation follows a rule you write ("when this happens, do that") and always does the same thing. Artificial intelligence interprets what it receives and decides how to respond, which is why it can get things wrong. If you can write the rule, use automation. If every case is different and arrives in free-form language, it is AI territory. And the best option is usually to combine them.

What the law says an AI is

The European Artificial Intelligence Regulation defines an "AI system" as a machine-based system designed to operate with varying levels of autonomy, that may adapt after deployment and that, for explicit or implicit objectives, infers from the input it receives how to generate outputs such as predictions, content, recommendations or decisions (Regulation (EU) 2024/1689, Article 3, point 1).

The key word is infers. An automation does not infer: it executes what you told it. An AI receives something it did not expect ("hey, do you have a slot on Thursday afternoon for something long?") and decides what to do with it.

The difference, in a table

	Automation	Artificial intelligence
How it decides	Follows fixed rules that you write	Interprets what it receives and generates a response
Result	Always the same for the same input	It can vary and it can be wrong
Best for	Repeated, predictable tasks	Free-form language, a different case every time
Control	Total: you can see the rule	You need to set limits and review
Example	When an invoice is paid, the thank-you email goes out	Understanding a WhatsApp message and replying

Examples to help you decide

Imagine a physiotherapy clinic. Two situations:

• Reminding someone of their appointment the day before. Always the same, at the same time, with the same text. This is pure automation. Adding AI only adds risk.
• A patient writes at eleven at night: "my shoulder has hurt since yesterday, can you see me this week?" There is no rule that covers every possible message. An AI understands the question, looks at the free slots and proposes a time, and if it is something a professional should see, it hands the conversation to a person.

(An invented example to explain the idea, not a real customer.)

The best option: let them take turns

In practice the two complement each other. AI understands what arrives in a messy form; automation does what comes next with precision. A message arrives, the AI understands it is a request for an appointment, and the automation creates the booking, sends the confirmation and schedules the reminder. The AI proposes, the rule executes.

When not to use AI

• When a simple rule solves it. It is cheaper, more reliable and easier to explain.
• When a mistake would be serious and nobody reviews it: money, health, legal matters.
• When you cannot tell the customer they are talking to an AI. That last one is not optional: the Regulation requires AI systems intended to interact with people to inform them that they are talking to an AI, unless it is obvious from the context (Article 50, paragraph 1), and sets 2 August 2026 as the general date of application (Article 113). Check it against the version in force when you read this, in case the calendar changes.

How many people use it

According to Spain's National Statistics Institute (INE), 21.1% of companies in Spain with 10 or more employees and an internet connection use artificial intelligence (first quarter of 2025, accessed on 11 October 2026). Four out of five have yet to start, so you are not late, but it is not a future experiment either. Among companies with fewer than 10 employees, the INE gives 13.4% employing AI, 5.9 points more than a year earlier.

What we have built at BAI

I use BAI's receptionist, which is the AI that faces the customer. According to what we publish on the website, we use both things and keep them deliberately separate. You choose BAI's automations with rules (what triggers them, conditions, what happens next). And the AI, which is in the receptionist and in the team assistant, says it is an AI, only answers with the information you give it, does not make up what it does not know (it hands the conversation to you) and can be switched off with a switch. The assistant, moreover, shows you what it is going to do before changing anything and waits for your confirmation.

Even so, if a rule is enough for you, you do not need AI, and I am telling you so myself.

What to do now

Make a list of the five things you repeat in your business and mark which ones you could write as "when this happens, do that". Those are automation. The ones you cannot write that way, because every case is different, are candidates for AI and for human review. To start with the first group, see what it means to automate a business and where to start.

How we checked this

I read the definition and the articles of the AI Regulation in its published text on EUR-Lex on 11 October 2026; the INE figure, in its press release. What I say about BAI comes from the published pages of our website. I wrote it with the help of artificial intelligence from the cited sources and from what we use at BAI; I sign it myself.

Sources

• Regulation (EU) 2024/1689 (Artificial Intelligence Act), Articles 3.1, 50.1 and 113: https://eur-lex.europa.eu/eli/reg/2024/1689/oj/spa (accessed on 11 October 2026).
• INE, "Encuesta sobre el uso de TIC y del comercio electrónico en las empresas. Año 2024 - Primer trimestre 2025": https://www.ine.es/dyngs/Prensa/ETICCE20241T2025.htm (accessed on 11 October 2026).
• Description of BAI's assistant and automations: published pages AI assistant and Ready-made automations.