Ever been in a shop where the staff kept recommending products before they’d even asked what you need?
You walk in for something simple — and suddenly you’re watching a demo you didn’t ask for, being told this product will “transform the way you work.”
Your AI vendor calls work exactly the same way.
They’ll sound like they’re learning your business — but really, they’re shaping your needs around their product. If you don’t know how to push back, you can end up spending significant money on something that solves the wrong problem beautifully.
How a £50,000 decision came down to one question
I worked with a healthcare organisation that was about to spend £50,000 on an AI tool. The demo was impressive. Slick dashboards. Big promises about efficiency gains.
But something didn’t feel right.
So before they signed, I gave them five questions to ask the vendor in their next call. The vendor couldn’t answer the first one — and the organisation walked away. They saved themselves £50,000 and what would almost certainly have been a year of frustration.
Here are the five questions. Save them before your next vendor call. They work for any AI tool, platform, or consultancy — including, for the avoidance of doubt, us.
Question 1: “Walk me through how you’ll get to know my business before recommending anything.”
A good answer: they describe a discovery process — questions they ask, how they observe your current operations, how they understand your goals before making any recommendations.
A bad answer: they jump straight into the demo. Or they describe the tool and tell you how it fits your business based on a five-minute conversation.
This question reveals immediately whether you’re speaking to a partner or a salesperson. A partner starts with your problem. A salesperson starts with their product.
This is the question the vendor couldn’t answer. They wanted to book the next demo call. They had no discovery process at all.
Question 2: “What’s the real 12-month cost — setup, training, the lot?”
A good answer: a clear breakdown of all costs — implementation, onboarding, training, integrations, support, and the monthly or annual licence.
A bad answer: they quote the monthly subscription cost and don’t mention anything else. Or they say “we’ll get to that in the proposal.”
The headline price is almost never the real cost. Implementation, data migration, staff training, and system integration routinely add 50–200% to the licence cost. You need to know the full picture before you can evaluate whether the investment makes sense.
Question 3: “Show me how this fits the way we actually work.”
A good answer: they ask to understand your current workflow first, then map the tool to it and explain specifically where and how it integrates.
A bad answer: they show you a generic use case diagram and tell you it “works for businesses like yours.”
Generic demos are designed to look applicable to everyone. What you need to see is whether the tool actually fits your specific process — not a stylised version of what your process might look like in a marketing presentation.
Question 4: “Can we try this on a small scale before committing?”
A good answer: yes — they can define a pilot, with clear metrics for success and a defined evaluation period.
A bad answer: they require a 12-month contract upfront. Or they offer a 14-day free trial that’s too short to evaluate anything meaningful.
Any vendor confident in their product should be able to design a pilot that lets you see real results before full commitment. If they won’t — or can’t — that’s worth noting.
Question 5: “What happens when the AI gets it wrong?”
A good answer: they describe specific failure modes they’ve observed, how the system flags uncertainty, what human review processes exist, and how errors are corrected and learned from.
A bad answer: “It’s 99% accurate.”
AI tools get things wrong. The question is not whether they will — it’s whether the vendor has thought carefully about what happens when they do. A vendor who hasn’t thought through failure modes hasn’t thought through responsible AI implementation.
The underlying principle
A vendor who starts with your problem is a partner. One who starts with their product is a salesperson. These questions are designed to surface which one you’re talking to — before you’ve signed anything.
They also apply to any consultant you hire. At Think Beyond Automation, the first thing we do before recommending anything is understand your business, your processes, and your actual goals. You can see exactly how we approach that in our AI strategy work.
If you’re currently evaluating AI tools or consultancies and want a second perspective, a free strategy call is a good place to start.
Know your real problem before any vendor call. That’s what separates good AI investments from expensive mistakes.