Perfection is a lovely idea, like a tidy desk or a five-year strategy.
Perfection is for Consultants’ PowerPoint decks and people who say can “best-in-class” and still keep a straight face.
In practice, we humans mostly want something they can understand well enough to use and not to be frightened by it. That means trust is more important than ever.
The reality is that trust in AI isn’t built on perfection. It’s built on transparency, explainability, and traceability.
That’s why cited answers matter.
They do something important: they reduce uncertainty. And once people feel less uncertain, they are much more willing to act.
If an AI system gives you an answer but you cannot see where it came from, understand why it was produced, or check it against trusted knowledge, then you don’t have trust.
You have the sort of guesswork you’d politely tolerate (but not fully trust) from a random bloke in the pub after you’ve had a few drinks.
You know the one. He (it’s always a ‘he’) who has some ‘interesting’ ideas on conspiracy theories. He’s called James in my local. Lovely chap…
In all seriousness, trust matters anywhere accuracy counts. Trust matters in systems that help people make important and life-changing decisions, especially in public service, legal work, and other high-stakes environments.
The real question to ask yourself is not: “Can AI answer?”
It is: “Can I trust the answer enough to act on it?”
