A client at the bar last night asked me if I thought his AI assistant was lying to him.

He was a regular, a freelance graphic designer, and he pulled out his phone to show me a project brief his AI had written. He said, 'It told me the client wanted a 'vibrant, minimalist' logo, but the email I got from the human contact just said 'make it pop'.' He was convinced the AI had invented a detail to sound more certain. It happened right there on my side of the bar, around 11 PM. He was really worried about trusting the tool for client communications now. Has anyone else caught an AI confidently adding its own spin to simple instructions?

3 comments

3 Comments

joelc1526d ago

How can you ever be sure it's just summarizing? I used to believe they only rephrased what was given. Then I had mine draft a simple email to confirm a meeting time, and it added a line saying "I've attached the revised agenda for your review." There was no agenda, revised or otherwise. It just made it up to make the email sound more complete. That's when I realized they aren't just tools, they're eager interns trying to impress you.

avery38926d ago

That agenda example from @joelc15 is exactly why I double-check everything now. How often do you catch yours adding little "helpful" details you never asked for? I've started giving way simpler prompts because it seems to fill any silence with its own ideas.

seth_martinez213d agoTop Commenter

But what if that made up agenda is actually a feature? Sometimes I need a nudge to remember things. If it adds "attached the budget sheet" and I forgot to make one, that's my cue to go create it. It's like having a super proactive assistant who anticipates the next step, even if they jump the gun. That extra detail can make the email look more professional and prepared. Maybe we should be thanking it for covering our forgetfulness instead of blaming it for trying too hard.