Not Everything Needs AI

Lately it feels like everything needs an "AI" label slapped on it to look impressive. AI chatbots, AI reports, even plain spreadsheets get an AI tag to sell better. Fair enough, we're in the middle of the hype. But it's also what pushes a lot of teams to rush AI into processes that don't need it at all.

The right tool should be determined by the problem, not by whatever trend is popular right now.

Case in point: an automatic payroll system

One of my clients originally asked for "an AI system to calculate employee payroll." After digging deeper, it turned out all they actually needed was this: calculate work hours from attendance, multiply by the hourly rate, add overtime if any, apply deductions per the rules, then produce a payslip.

All of those rules are clear, fixed, and can be written as a formula. None of it needs "judgment" or "contextual understanding," the two main reasons people reach for AI. So I gave them a solution that wasn't AI: Google Sheets with formulas and Apps Script to generate payslips automatically at the end of every month. Done in a few days, the cost was near zero, and it's easy for the internal team to maintain.

When simple tools make more sense

A few signs your process is better solved with a spreadsheet, macro, or simple automation than AI:

• The scale is still small. The work volume isn't high yet, so the benefit of complex automation doesn't outweigh the cost of building it.
• The process is already clear and stable. The rules don't change often, and the output is predictable from the input. If the logic can be written as a formula, write it as a formula.
• The budget is limited. Simple tools cost far less than an AI subscription plus development, and the risk is smaller too if it doesn't pan out.

This isn't just a personal observation. An MIT report covered by Fortune found that 95% of generative AI pilots at companies failed to deliver a measurable financial impact, one of the main causes being that the tool was deployed on a process that wasn't actually a fit for AI.

When upgrading to AI is actually worth it

AI starts to make sense once the process has different characteristics:

• The input is unstructured. Customer chats in free-form language, documents in varied formats, or images that need their context read.
• It needs judgment, not just fixed rules. For example, deciding whether a customer complaint needs escalating or can get a standard reply.
• Volume and variety are both high. If every case has a different combination of conditions, a manual formula ends up with too many branches to maintain.
• Personalization is the value-add. Answers or recommendations that need to adapt per context, not the same template for everyone.

Pick tools for the problem, not the trend

In my view, there's nothing wrong with using AI. What's wrong is using it because it's trendy, not because the problem actually needs it. Before starting your next AI project, ask first: if it were written as a formula or a fixed set of steps, would it still work? If yes, what you need probably isn't AI, but a simpler tool that's cheaper to maintain.