Enhance AI Trust: Designing for Mistakes, Not Just AccuracyEnhance AI Trust: Designing for Mistakes, Not Just Accuracy
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
"The AI feature is ready. Can you make it look nice?"
That's how the brief started.
By the end of the call, we'd quietly redesigned the whole feature — without touching a single screen.
Here's what actually happened on that call — and the lesson for anyone shipping AI features.
I asked one question: "What does the user see when the AI gets it wrong?"
Silence.
The team had spent months on the model. Accuracy was impressive. The demo was smooth.
But nobody had designed the moment the answer is wrong — and the user knows it.
So instead of "making it look nice", we worked through:
→ The wrong-answer moment. The user spots a bad recommendation. One tap to correct it — or lose them. We designed the correction first, the happy path second.
→ The reasoning line. One sentence under every result: why the AI suggested this. Not the math — the logic a human can judge.
→ The honest empty state. Before the AI has enough data, it says so — instead of guessing confidently and burning trust on day one.
Three weeks later: same model, same accuracy.
But users stopped abandoning the feature after the first mistake.
The model didn't get smarter. The interface got more honest.
I've seen this pattern on more than one product now: "make it look nice" is never the real brief. The real brief is: make people trust it.
What does your AI feature show when it's wrong?:
Back to feed
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started