Most bad AI purchases come from asking "what can it do?" instead of "what problem does it solve for us, and at what real cost?" A short, sceptical checklist will save you more money than any demo. Vendors sell capability; you should buy outcomes.
The questions that matter
- What specific task will this replace or speed up? If the answer is vague, stop.
- What does it cost fully loaded? Licences, integration, training and oversight time.
- Where does our data go? Storage, retention, and whether it trains the vendor's model.
- How accurate is it, and how do we check? Ask for error rates, not adjectives.
- What happens when it is wrong? Who catches it, and what is the cost?
- How locked in are we? Can we export our data and switch later?
- Who owns the output? Legally and commercially.
Run a real test, not a demo
Vendor demos are staged. Insist on a short trial with your data and your task, measured against a baseline. This is the practical side of knowing how to tell which AI tools actually work.
Match the tool to the strategy
A tool should serve a decision you have already made about where AI helps - not the other way round. If you cannot connect it to a goal, it is a distraction, however impressive. Keep it tied to what goes in your AI strategy.
The bottom line
Good AI buying is disciplined questioning, not enthusiasm. Learning to evaluate tools like an investor - on value, risk and evidence - is a core skill in the Strategic Application of AI in Business course at London School of Business UK. Enquire today.