Does your software "have AI" or just a pile of disguised IF/ELSE?

It became fashionable to say software "has AI".

Open the sales deck of any customer service system today and you will find the badge: "with Artificial Intelligence". And as someone who builds these systems, I will be blunt: much of what is sold as AI is a rigid decision flow with a nice name on top.

It is not malice, it is marketing. AI sells, so everyone slaps on the label. The problem is that you pay the bill — buying expecting one thing and getting another.

The difference nobody explains

There is a huge gap between rule-based automation and an AI agent. Rule-based automation is the decision tree, the "press 1 for Billing", the bot that follows a fixed script (if/else). It works while the customer behaves exactly as the programmer predicted. An AI agent understands natural language, handles what was not in the script and decides what to do — including recognizing when to hand off to a human.

The hood test

Four questions reveal which one you are about to pay for: 1) What happens if the customer asks something unplanned? 2) Does the answer change with conversation context? 3) Can I update the knowledge without reprogramming the flow? 4) Does it know when to call a human and pass the context along? If the vendor only talks tech and never about these, be careful.

If you manage properties or run a security company, your support is dominated by repetitive questions. A rule system jams; an AI agent absorbs that volume and frees your team for what truly needs a person.

Continues in [Part 2](/en/blog/agente-de-ia-mas-cade-o-retorno). Originally published on LinkedIn.