AI is everywhere...what’s one thing it still can’t do right?
I work with startup teams and use AI tools almost every day for writing, research, support, and more. It saves a lot of time, no doubt. But there are still things AI just can’t do well.
For example, when reading customer feedback, AI often misses the real meaning behind the words, like when someone sounds polite but is actually unhappy. Or when making product decisions, AI can list pros and cons, but it doesn’t really understand the bigger picture or timing like a human would.
These things may seem small, but they make a big difference in how startups grow and build trust.
Little gaps, big impact.So I’m curious:
👉 What’s one thing you still can’t trust AI to do in your startup or day-to-day work, and why?
Drop your stories, rants, or quick thoughts below. Let’s map where human insight still rules.
Replies
I think it has become pretty good at many things. Still unable to create something for me/do for me in a physical world – like cleaning dirty plates and mess :D
@busmark_w_nika this happening too, but not so casually 😂
Make my damn coffee
@ross_danowitz Alexa?
@hamza_afzal_butt AI without context is just a chat bot. Remember, the AI we have today is the worst AI we'll ever have as it only gets better.
@martydIf this is the worst, I’m both excited and terrified to see version 2.0
One thing AI still can't do right is understand human context and emotion with real nuance.
@rahulparma2807 So true. It can talk the talk, but when it comes to reading between the lines... It’s still learning the language of being human 😅
@rahulparma2807 This is true, but on the other hand, more personalized and human-understanding products are emerging day by day. We are working on such a product too.
@rahulparma2807 @armutyus that's amazing. What is your product mainly about?
@rahulparma2807 @hamza_afzal_butt
Briefly it's a digital memory and personalized AI, i don't know if it's okay to share here but you can check it out from my profile.
@rahulparma2807 @armutyus Ahhh, I love the idea of having a digital brain that actually remembers things for you. Good job :)
FairPact AI
Even the best AI systems still have a lot of trouble parsing legalese. Contracts, laws, regulations...all of these things require a nuanced and informed understanding of so many things that AI tends to do poorly.
There are ways to use AI for small subtasks used in legal reasoning (we're about to release a product related to this in fact), but even the best systems aren't at the point where you can just drop a contract in and expect the AI to fully and properly understand every part of it without high risk of hallucination.
@john_licatoThat’s super interesting...and honestly a bit scary when you think about how much rides on legal interpretation. Your upcoming product sounds like it could be a real game-changer in that space. Would love to hear more when it’s live!
FairPact AI
@hamza_afzal_butt yeah I'm always scolding people when they tell me they're just putting legal documents into chatgpt, or asking it for legal advice, and just trusting what comes out of it uncritically!
@john_licato yep, I’ve seen that too ):
I have so many duplicate files and photos on my laptop. Can AI help identify and remove them?
@lucytaylor01 Lucy, there are tools like Gemini Photos (Mac/iOS) or Duplicate File Finder on Windows that use basic AI to spot duplicates. It's worth trying; just back up first in case the robot gets overzealous!
@daniel_james20 Totally get where you are coming from. The over-reliance is real, and when it fails, it fails hard. I think it's more of a tool than a replacement but yeah, it’s definitely not at the “trust blindly” stage yet.
Totally feel this! AI still stumbles hard on tone—like when a customer says ‘Great, thanks…’ but actually means ‘Fix this now or I’m out.’ 😂 Maybe it needs a crash course in passive-aggressiveness and MBTI memes before it cracks sarcasm.
As for changing the world… if AI could yeet dictators into a virtual trash bin, sign me up for beta testing! But for now, I’ll settle for it nailing basic empathy in emails. Baby steps should count, right?
(Still rooting for AI to at least auto-delete war propaganda though. A bot can dream.)
@rani_zagita This cracked me up! 😂 “Passive-aggressiveness and MBTI memes” needs to be part of AI training ASAP. And yes to empathy in emailssuch a small thing, but it changes everything.
To be honest, from a content writer's perspective, I feel that even though it provides me with content, it still feels generated rather than curated.
Why am I saying this? It is because I can still feel no emotions, opinions, or constructive reasoning that can impact my vision. I may use it to understand something myself. However, when it comes to making the reader apprehend the same while keeping engaged with the content, I would not trust AI with it.
That being said, trusting AI with content in itself, I still can't do it. It still gives off robotic vibes.
@diksha_singh15 So well said. AI can assist, but it can’t replicate that human spark, those subtle opinions and nuances that make content actually stick. You nailed it with “generated vs. curated.”
Intellectia.AI
Handle super-sensitive client negotiations
@feichen_ ahh, right!
Intellectia.AI
@hamza_afzal_butt Hi Hamza, we're going to launch our app on July 9 or 16, we need a hunter to help our launch, are you available?
SEO Stuff
AI greatly helps, but I still can’t trust it to write anything emotional. Yesterday I tried using for a support reply and it sounded like a robot giving a TED Talk 😅. AHH, some things just need a human touch.
@haiqa_irfan Right, we (humans) are still the guardians of real empathy. 😅