Why Some AI Users Get Results While Others Spin Their Wheels

The gap between AI power users and everyone else has less to do with the technology and more to do with how it is used, according to a new Fast Company guide on extracting value from chatbots. The piece makes a blunt claim: results depend heavily on technique, not just the model behind the curtain.

The guide identifies five levers users can pull. None involves buying anything. All involve changing behavior.

Does prompt precision actually change the outcome?

Yes, the guide argues, and by a wide margin. Users developed an implicit feel for search engines over decades — Boolean operators, precise proper nouns, better queries. The same skill transfers to chatbots. Specificity about what information you want makes a measurable difference, particularly when AI is used as a search tool.

The guide offers a concrete example. Shopping for a bookshelf no longer means sifting through listings. Instead, the author photographs the spot, supplies the dimensions or range of dimensions, and asks the chatbot to find the best options within those constraints. The payoff is saved time.

One adage anchors the section: "The more you know, the more they know." The chatbot should be treated as an earnest research assistant with the capacity to make a lot of mistakes — not as a mind reader. That framing sets the tone for everything that follows.

What about the mistakes themselves?

Here the guide is at its most skeptical, and Business Bearings readers should take note. The author's advice is blunt: "Really, don't do this!" — meaning, do not offload all tasks to ChatGPT or Claude.

The recommended defenses are procedural, not technological:

• Check any chatbot-produced work yourself, first and foremost.
• If settings allow, instruct the AI to be more humble and cautious when making claims.
• Tell it to consider multiple ways of thinking about a problem and present divergent viewpoints.
• Tell it to admit when it is unsure.
• Always demand direct citations — and verify them off the platform, not on it.

The guide warns that a chatbot claiming to be "double-checking" something does not mean it actually is. AI can cite unreliable sources and hallucinate citations entirely, generating fake web pages and then apologizing when they turn out not to exist. The recommended mental model: a sieve that helps you sift through information faster, not a trustworthy judge of quality. Never treat it as an objective source of information.

The stakes are cognitive as well. The guide draws a parallel to calculators: some people's mental math atrophied through reliance, while others used the tool to tackle more advanced math. Offload your thinking to AI and you risk dulling creative problem-solving and growing dependent on a flawed crutch; use it to stress-test your work and outsource routine tasks, and it feels like an upgrade.

Is your chatbot remembering more than you think?

It is, the guide confirms. AI platforms retain what you ask and what you tell them — a practice common across tech, usually in service of tailored ads. Users who object should check privacy settings or avoid AI entirely.

For those comfortable with retention, memory becomes a tool. Tell the AI to always introduce you the way you prefer in drafted emails, to remember that you like offering Zoom calls, or to keep a formal or casual tone. Shoppers can instruct it to stop surfacing Amazon links, or to prioritize New York Times links they can actually access with an existing subscription.

A caveat follows the promise: asking the AI to erase something from memory carries "absolutely no guarantee" it will comply. Check privacy and memory settings independently.

Can you shape how AI talks back?

Yes, and the guide treats this as an underused advantage over Google. Users can demand bullet points instead of essays, request visuals and graphics, or receive answers as Python code to run deterministically and verify the model's work. Bilingual responses shown side by side can even build fluency in a second language.

The author's own customization pushes against homogenization. Citing a "somewhat well-founded fear" that AI is flattening language, the guide describes instructing the chatbot to use unusual, niche vocabulary — and supplying a list of common AI crutch words to avoid.

What are tokens, and why do they matter?

AI runs on tokens: input tokens for prompts and context, output tokens for responses. Long prompts burn more input tokens; long responses burn more output ones. Users may not see the count, but usage limits make it felt.

The fix is explicit instruction — tell the AI to be brief when one link suffices, and to go deep when detail is warranted. The guide's quiet implication: metered consumption is a skill, and unmetered use eventually costs you.

The through-line is control. The technology, the guide concludes, is what you make of it — and the users winning with AI are the ones who never forgot that.