Read a sample: 1. Why the Words Matter
On a Tuesday morning, Maya sees three claims about technology before she has finished her coffee.
A package for a new thermostat promises a smart home. A shopping app says its suggestions are powered by AI. At work, a message announces that a chatbot can help the team find answers faster.
None of these claims is necessarily false. Each might describe a useful product. But each leaves Maya with questions the label alone cannot answer.
Will the thermostat adjust itself, or will it simply follow a schedule she enters? Are the shopping suggestions based on patterns learned from customer activity, a list of popular items, or something else? Does the workplace chatbot search company documents, generate answers from a language model, or combine the two?
Maya does not need to become a programmer to ask those questions. She needs a way to separate the words used to sell a feature from what the feature actually does. That is what this book is for.
Artificial intelligence has become an everyday phrase, but the things gathered under it are not all alike. A program that sorts photographs, a tool that writes a paragraph, and a system that helps predict equipment problems may all be described as AI. Meanwhile, an ordinary timer may be called “smart” even if it does nothing more than follow a rule. A product can also combine several methods: some automated steps, some machine learning, and perhaps a generative AI feature.
These distinctions are not just matters of technical etiquette. They affect what you can reasonably expect from a product, what kind of mistake it might make, and what questions you should ask before relying on it.
Imagine that Maya opens the thermostat box. The front says “smart temperature control,” but the instructions tell her that she can choose a temperature for weekday mornings and another for evenings. At the chosen times, the thermostat changes the setting.
That is useful automation: a device performs an action according to a schedule. The schedule may save Maya from making the same adjustment every day. It does not, by itself, show that the thermostat learns from her behavior or uses artificial intelligence.
Now imagine a different feature in the same product. The instructions say the thermostat can use information about past adjustments to suggest a schedule. That is a different claim. It raises different questions: What information does the feature use? Does Maya approve the suggested schedule? Can she turn the feature off?
Even then, the wording may not reveal the precise method. A suggestion based on past adjustments could be produced in more than one way. The product might use a machine learning model, a simpler calculation, or a combination of methods. Unless the company explains enough, Maya may not be able to tell.
That is an important result, not a failure. Sometimes the best answer is “I don’t know from the information available.”
The same caution applies to familiar words such as smart, intelligent, personalized, and AI-powered. They can point toward a feature worth investigating, but they do not provide a complete technical description. “Smart” might mean that a device connects to a phone. “Personalized” might mean that a service uses a preference Maya selected. “AI-powered” might refer to one part of a larger product rather than every action the product performs.
A label can invite a question. It cannot answer every question on its own.
When Maya encounters a technology claim, she can examine it in three layers: the name, the behavior, and the method.
The name is what the company, headline, or coworker calls the feature. It might be “smart scheduling,” “AI recommendations,” or “your virtual assistant.” Names are useful shorthand, but they are often broad. Two products with similar names may work differently.
The behavior is what the feature observably does. Does it change a temperature at a set time? Offer three products? Produce a written answer? The behavior is usually easier to investigate than the method. Maya can try the feature, read an explanation, or look for a demonstration.
The method is how the system produces that behavior. Does it follow rules written by people? Does it use a model developed through machine learning? Does it retrieve existing information, generate new wording, or do both? The method may be described in product documentation, but it is not always visible from the outside.
These layers help prevent two common mistakes. The first is assuming that a feature uses AI because its name sounds intelligent. The second is assuming that Maya can identify the method just by watching the result.
Suppose a shopping app recommends a raincoat after Maya looks at umbrellas. The behavior is clear enough: the app displays a raincoat suggestion. The reason is not. A person may have created a rule linking umbrellas and raincoats. The app may use patterns from shopping activity. It may combine several signals and methods. Maya can observe the recommendation without knowing which explanation is correct.
The distinction also works in the other direction. If a company says it uses machine learning, Maya still needs to know what the feature does. A technically accurate description does not tell her whether the suggestion is useful, whether she can change it, or whether the product fits her needs.
A practical first pass is to ask:
- What is it called? Notice the label without treating it as a complete explanation. - What…