When not to use AI (and what to use instead)

AI is now the default answer to every business problem, which is a good reason to be careful. A language model is the right tool for some jobs and an expensive, unreliable one for others. Choosing well saves money and avoids systems that quietly get things wrong.

A quick test

Ask three questions. Could you write the rule down exactly? Is the input always in the same shape? Does the answer have to be right every time? If the answer to all three is yes, you probably do not need AI. You need ordinary software.

Use something simpler when

• The logic is fixed. A formula, a threshold or an if-then rule is cheaper, faster and fully predictable.
• The data is structured. A database query or a report does it without guessing.
• It runs on a schedule. A scheduled job, not a model, sends the weekly email.
• Mistakes are costly and unreviewable, such as tax computation or payments. These need deterministic code and tests.
• A form would do. If you only need information in a fixed shape, ask for it in that shape.

AI earns its place when

• The input is messy: free-text emails, documents, voice, images.
• The task needs judgement across context, such as classifying an enquiry or summarising a thread.
• The volume is high enough that doing it by hand does not scale.
• A wrong answer is recoverable, or a person reviews the uncertain cases.

The best systems use both

In practice the strongest setups combine them: a model reads the messy input and turns it into structured data, then ordinary code applies your rules, updates your systems and decides who should review. The model handles the unclear part. Everything else stays predictable and testable.

Not sure which side of the line your problem sits on? Describe it in a few sentences, or read where to start with AI automation.