What's something AI is actually terrible at that nobody talks about?
I'll take the hit.
AI has no idea when someone is politely furious.
You know the email. "Hi team, just circling back on this again as I haven't heard anything. Thanks for your attention to this matter." Reads like a sweet grandma wrote it.
A human reads that and thinks "oh no, they are about to burn the building down." AI reads it and thinks "great sentiment, very positive, 98% satisfaction score."
I tested this with actual support tickets. Customer writes "I'm sure you're very busy, but it's been three weeks and my account is still locked. No rush though!" AI gave it a 9 out of 10. The customer cancelled the next day.
AI also can't tell when someone is lying in a meeting. You know that person who says "love that idea" with dead eyes? AI believes them. Every time.
Or when someone says "let me think about it" and you never hear from them again. AI marks it as "promising lead."
It reads what you type, not what you mean. Because it's never been ghosted. It's never had someone say "we'll circle back" and then disappear forever.
What's the thing AI completely misses that drives you insane?
Imed Radhouani
Founder & CTO – Rankfender
Building what you ask for, not what I think is cool
Replies
sarcasm detection is the one that gets me. founders write "sure, another todo app, that's definitely what the world needs" and AI reads it as genuine enthusiasm. spent way too long figuring out how to handle that edge case. turns out ironic founders are the hardest input to get right.
@nerdkick11 I relate to this so much. When I read sarcastic lines like that, I instantly get the meaning but AI takes it literally. I think I’d try layering sentiment with contradiction detection. Still, I agree, founders who joke like this are tough inputs to handle properly.
@dalhat_usman
Contradiction detection is interesting — hadn't thought about layering it that way. right now i'm just flagging when the sentiment score doesn't match the word choice, but it's crude. The sarcastic founder problem might be unsolvable with text alone honestly.
Rankfender
@nerdkick11 Oh man the sarcasm thing is brutal. "Another todo app, just what we needed" is clearly a roast. AI reads it as "this user is expressing satisfaction with the product category."
The ironic founder is the hardest user persona. You can't just ask them what they think. You have to decode what they actually mean. AI doesn't stand a chance.
@imed_radhouani
Exactly. The ironic founder writes "great, another AI tool" and means "i hate this." the earnest founder writes the same thing and means it. Zero difference in text. Context is everything and AI has none of it.
@imed_radhouani this hits hard :) AI also completely misses when someone says "interesting" about your product pitch. That single word response? AI thinks they're engaged.
Rankfender
@rohanrecommends Ha yes. "Interesting" is the most dangerous word. AI reads it as "this person is intrigued." Any human knows it means "I have no idea what to say and I'm trying to be polite."
Same with "we'll circle back" and "let me think about it." AI logs it as a promising lead. You know you're never hearing from them again.
Honestly, I’ve noticed the same thing. I rely a lot on gut feeling when reading messages and AI just doesn’t have that. For me, it’s obvious when “no rush” actually means “this is urgent” but AI takes it literally every single time.
Rankfender
@dalhat_usman The "no rush" thing is the classic. You hear that and you know it means "this should have been done yesterday." AI just nods and puts it in the low priority pile.
Same with "just checking in." AI logs it as routine. Human knows it means "I've been waiting too long and I'm starting to get annoyed."
It's the stuff between the words. AI can't read it because it's not there. But it's the only part that actually matters.
Spot on, those "circling back" emails are pure passive fire, and AI's blind spot there kills me every time.
For me, it's AI missing micro-lies in sales calls: the "Sounds perfect, send over the proposal" with zero follow-up intent. Humans catch the flat tone or hesitation; AI scores it green and wastes everyone's cycle.
Rankfender
@swati_paliwal Yes. The "sounds perfect" with flat tone is the one. AI hears the words and marks it as a hot lead. Any human knows they've already moved on.
We had a sales call where the guy said "this looks great" three times in a monotone. AI gave it a 92% satisfaction score. He ghosted us the next day.
The worst part is you can't train it on tone. There's no data for "polite but completely uninterested." So it just keeps happening.