The hype tax
Every few years a technology arrives that seems to promise everything. Right now it is AI, and the pressure on business owners is real: competitors are announcing AI initiatives, vendors are adding “AI-powered” to every product page, and it can feel like standing still is the riskiest move of all.
But adopting a tool because it is fashionable carries a cost we call the hype tax. It shows up as projects that take longer than planned, systems that behave unpredictably, monthly usage bills nobody budgeted for, and staff who quietly go back to the old spreadsheet because the new thing is harder to trust.
The goal was never to “use AI.” The goal is to reduce cost, increase throughput, and free your people for higher-value work. Sometimes AI is the best way to get there. Often, it is not.
Where plain automation wins
Traditional automation means software that follows explicit rules: if this, then that. It is not glamorous, but it has properties that matter enormously in day-to-day operations.
- It is predictable. The same input produces the same output every time, which makes it easy to test, audit, and trust.
- It is inexpensive to run. Once built, a scheduled script or integration costs very little to operate.
- It is explainable. When something goes wrong, you can trace exactly why.
Moving order data from your e-commerce platform into your accounting system, generating invoices from a fixed template, syncing inventory between two systems overnight, or sending reminders based on dates in a database: these are rule-based problems. Adding a language model to them introduces variability where you want none.
If you can write the rule down, you probably don’t need AI to follow it.
Where AI earns its place
AI shines where rules break down: when information arrives as language rather than fields. Reading a free-form customer email to understand what the person actually wants, summarizing a long report, pulling key terms out of contracts that are all formatted differently, or drafting a first-pass response for a person to review.
In these cases, writing rules for every variation would be impossible, and a well-designed AI step can save real hours. The key words are well-designed: the AI should handle the interpretation, while dependable logic and people handle the decisions that carry risk.
A simple test before you spend
Before choosing a tool, ask a few plain questions about the task itself:
| Question | Points toward automation | Points toward AI |
|---|---|---|
| What does the input look like? | Structured fields, fixed formats | Free text, varied documents |
| Can the steps be written as rules? | Yes, with few exceptions | Not without endless exceptions |
| How costly is a wrong answer? | High; needs to be exact | Moderate; a person reviews |
| Does the output need to be identical each time? | Yes | No, it just needs to be useful |
If most of your answers land in the left column, start with conventional automation. If they land on the right, AI may be worth a pilot. If they are split, you are likely looking at a combined approach, which we cover in our field guide.
The takeaway
The smartest technology decision is the one that solves the problem reliably at the lowest total cost. That might be an AI model. It might be a fifty-line script. It might be a process change that needs no software at all.
We spent decades solving business problems with technology before AI became fashionable. That experience is exactly why we don’t reach for it first.
