A pitch is marketing, not a technical spec
This isn't a knock on vendors – selling is part of the business. But as a decision-maker, you need your own follow-up questions to find out what's really behind the claims.
Four follow-up questions that cut through any pitch
What exactly is the system allowed to do on its own?
Don't ask "Can it do X?" – ask "What happens, concretely, step by step, when a customer writes Y?" Concrete workflows expose vague promises (see also module 10, AI agent).
Which real reference customers, with what documented result?
"Used by leading companies" without a name or a number is a claim, not a reference. Ask for a concrete, verifiable case study.
What happens when the system gets it wrong?
Every AI system makes mistakes occasionally (see module 4, limits and risks). The key question isn't whether, but what happens then – is there an oversight mechanism, an approval step, a log?
What does it really cost, including operations?
A low entry price can hide ongoing maintenance, support, or extra usage fees. Always ask for the total cost of ownership over, say, 2-3 years (see module 9, cost & ROI).
Why this matters for you as a decision-maker
A vendor who can answer these four questions concretely and without dodging probably has a solid system. Evasive or purely promotional answers are themselves the clearest warning sign.