How to Read an AI Vendor Pitch Without Getting Fooled

Somewhere this quarter, you will sit through several AI vendor pitches. They will all say roughly the same thing: this tool saves your team hours, cuts costs, pays for itself within weeks. The slides look sharp. The demo runs flawlessly, because it is running on the vendor's best case data, not yours.

The AI tools market has exploded over the past year. Every SaaS product now has an AI feature on its homepage, often bolted onto something that already existed. Leaders approving budgets are being asked to evaluate technology they don't fully understand, based on a thirty minute demo built specifically to convince. In the workshops we run at MAKIA Labs, this comes up in nearly every session with leadership teams: someone has just walked out of their third demo of the month, convinced, and unable to explain exactly why. The real gap isn't technical. It's a missing method for evaluating what you're being shown.

Most leaders judge a pitch by asking whether the tool works. The demo will always say yes, that is its only job. The question that actually matters is different: does this tool fit into a workflow that already exists, with people who already exist, doing work that is already defined. A tool that works in a vacuum and a tool that fits your actual operation are two different things, and no demo can answer the second question.

Four questions worth asking before you sign anything:

1. Ask for the failure mode, not the success story. A good vendor can tell you exactly when their tool doesn't work: for which kind of data, which kind of team, at what volume. If they can't name a single limitation, either they haven't tested it under real conditions, or they aren't being honest with you. Both are worth noting.
2. Ask who on your team will actually maintain this tool six months from now. Not who presented it in the sales call, not who signed off on it: who will configure it, fix the errors, train the next new hire. Adoption almost always dies in the gap between the person who buys the tool and the person who has to use it every day.
3. Ask what happens if you stop paying tomorrow. Data portability, vendor lock in, what actually breaks in your operation if the company disappears, gets acquired, or triples its price at the next renewal. A good answer is specific. A vague, reassuring answer is not an answer.
4. Ask for a reference customer your size, in your industry, not their biggest logo. A case study from a three thousand person enterprise tells you nothing about what will happen on a forty person team running different processes.

Before any of this, the more basic question is one of meaning: what problem are we actually trying to solve, and would we still want this outcome if AI weren't part of the equation. If a leader can't answer that first, no amount of vendor due diligence will save the decision. There is also a question of actors: the people who will live with the tool daily should be in the room before the contract is signed, not informed afterward.

A good vendor pitch doesn't need you to fall for it. It needs you to ask good questions and still say yes. That is a much higher bar, and the vendors worth working with clear it every time, without flinching.

Try this week: before your next vendor call, or before you finalize a decision already in progress with a current vendor, ask them directly: "Tell me about a customer where this didn't work, and why." Write down the answer word for word. If they don't have one, that's an answer too.