The fastest way to waste money on AI is to start with the tool. The cheapest way to make money from AI is to start with the question. Before you sign up for a subscription, or commission a build, walk through these five.
1. What specific task am I trying to remove or shorten?
"Use AI" isn't a goal. "Stop spending two hours every Friday formatting the weekly report" is. If you can't name the task in one sentence, you're not ready to buy a tool, and you're definitely not ready to commission a build.
2. How long does this task take me today, in hours per week?
Be honest. If it's under an hour a week, automating it probably isn't worth the setup. If it's a day a week across the team, you've found something worth doing properly, and quite possibly something worth building a small internal tool around.
3. What happens if the AI gets it wrong?
For drafting a reply, the answer is "I edit it". For filing your VAT return, the answer is "I get fined". The bigger the consequence, the more human review you need in the loop. A good build bakes that safety net in; a bad build hides it.
4. Where will the data live, and who can see it?
If you're pasting customer information into a free chatbot, you've effectively shared it with a third party. A custom build can keep that data inside your own account and out of model training. Either way, ask the question on purpose, don't stumble into the answer.
5. Who in my team will actually use this on Monday?
A tool nobody opens is worse than no tool, because you paid for it. Before buying anything, or commissioning anything, name the person, the moment in their week, and the result that proves it's working. Every build I deliver includes a short training session for exactly this reason.
If you can answer all five clearly, you're ready to spend money. If you can't, the next step is a conversation, not a purchase. That's what the ad-hoc advice call is for.