When Not to Use AI

Should you use generative AI for everything?

AI can write code, design presentations and analyse data so well that it is easy to think it can do every task. But when you need certainty, it is the wrong tool.

Say your company pays monthly incentives to 3,000 sales staff, each with a different target, slab and cap. Every incentive must be exactly right.

Generative AI does not calculate, it predicts. Across 3,000 rows, it can skip a few, mix up two names or apply the wrong slab. Run it again next month, and it may handle the same case differently.

What should you use instead?

An Excel formula, or ask AI to prepare a dashboard using code, so the code does the calculation, not AI's guess.

If AI writes the code, is the result always right?

Only if the code is right. If it picks the wrong column or misses a slab, every number will be wrong. Test it with a few cases you already know.

When is it fine to let AI answer directly?

Ask yourself: would a wrong answer cost money, a customer or a decision? If not, like drafting an email or brainstorming ideas, let AI answer. If yes, use a formula or code.

What to remember

When you need certainty, do not rely on AI's answer. Let AI build the formula or code, test it with numbers you know, and then trust the tool, not the guess.