What’s Wrong with AI for Support?
AI in customer support is booming - but I’m curious, what are the biggest pain points you’ve seen with the solutions you’ve already tried? and what were those solutions?
here are my top3:
- most AI agents for support feel too generic - answers are painfully dry, and they often hallucinate. I’ve heard a feedback where the bot “confirms” a refund request but never actually does it.
- AI pricing is a black box - is it per resolution, per token, per feature???? no clue, you even end up paying for hallucinations. It’s expensive, locked into yearly contracts, and hard to predict.
- setup bottlenecks - to change anything or set up new data, you have to go through the support team… who might take ages to respond. I want to load a new knowledge base now and see it working instantly.
soooo,
what's your top3?
Replies
Scade.pro
I think it depends on the quality of the data. I see how products/stores are becoming more aware of it, and how the quality of replies and suggestions is getting better and better. This leads to the greater conversion. I can see this as a customer myself. But tbh I hate voice assistants in government institutions, they are just so slow and dull :(
@nastassia_k agree, voice assistants is a pain! I'm not against basic instructions, like 'press 5 if you need to do this and that', etc, but not when they are trying to emulate a real person. I don't even doubt that it's a question of time, but I'm still not sure I want this to happen, tbh 😅
IXORD
I think that we need a couple of levels for customer support, level 1 for those who have general questions. And the second level where a real person will understand the client's problem. I think such systems already exist
CoSupport AI
@ixord that’s exactly the gap we’ve been working to close with our own AI support agent — one that handles up to 90% of repeat tickets with 99% accuracy and zero hallucinations. we just launched it on PH if you’re curious to see how it works https://www.producthunt.com/products/cosupport-ai
AI for support feels wrong because there's no one to take responsibility and no one is accountable. Humans are wired to seek help from other humans - and we're okay if they're using tools. However, with AI - we know that we're talking to a trained machine.
CoSupport AI
@thebigk also that question’s been at the core of building our own AI support agent – it’s designed to give instant, accurate answers without hallucinations, and still feel human in tone. just launched it on PH if you want to see how close we’ve come https://www.producthunt.com/products/cosupport-ai
I'd say it's wrong to rely on AI solely. It's a fantastic tool that helps you assist customers much faster and sometimes even better, but we've got to double-check what AI suggests.
This way, you don't have to spend time on creating your response, but only alter the one AI made, which is already really helpful.
Oof, totally feel this. My top 3?
1 Bots that sound like a corporate intern on autopilot – over-polite but useless.
2 Hidden costs everywhere – feels like playing Minesweeper with my wallet.
3 ‘Plug-and-play’ that’s actually ‘plug, wait 3 weeks, then maybe play.
My issue? Companies use AI support as an excuse to fire their actual support team, then act surprised when customers hate it. It's cost-cutting disguised as "innovation" 😂