← All posts

The cheap-AI trap: when free tools quietly cost you more

There's a cheaper AI option for everything now, and cheaper is often the smart call. But sometimes "free" is the most expensive line in the business. Here's how to tell.

New AI tools launch every week promising to do it cheaper, or free. A lot of the time that's genuinely good news — the cost of capable AI is falling fast, and I use cheaper options myself where they're good enough. But "free" has a way of hiding its price somewhere you're not looking. Here's how to spot when the cheap route is actually the expensive one.

When cheap is exactly right

For general, low-stakes, human-checked work, spend as little as you can. Drafting, summarising, tidying up text, getting unstuck. A free or budget tool does this fine, and paying more buys you very little. If a person eyes every output and nothing critical rides on it, cheap is the smart, disciplined choice. Don't let anyone upsell you here.

Where free quietly bites

Your data becomes the payment. If you're not paying, sometimes what you type is the product. For a business handling customer information — especially anything regulated — that's not a saving, it's a liability. Read what a free tool does with your inputs before it touches a single customer detail.

The time tax. A cheaper tool that's a bit worse means someone spends longer fixing its output, every day, forever. Multiply a few wasted minutes across a team across a year and the "free" tool has cost more in wages than a paid one ever would.

No reliability, no support. Free tools vanish, change overnight, or break with nobody to call. Fine for a side task. Not fine for something your team depends on to get through the day.

The integration gap. A cheap tool that doesn't connect to your systems leaves your people copying and pasting between tabs. You saved on the licence and paid it straight back in admin.

A simple way to choose

Match the spend to the stakes. Low stakes, human-checked, occasional — go as cheap as it works. Core to how your team operates every day, touching customer data, or something a mistake on would be costly — pay for the tool that's reliable, private and actually connects to your business. The mistake isn't spending too little or too much in general. It's spending like it's low-stakes on something that quietly turned out to be critical.

The test before you commit

Before you standardise on any tool, free or paid, run it on a real job for a week and count the time spent around it, not just in it. The true cost is the licence plus the workflow. Judge on that number and you'll rarely get the cheap-versus-worth-it call wrong.