Did we really let customer service get so bad that people think this will be an improvement? My gut feeling says this will not transform companies who are bad at customer service to improve, rather the ones that are already bad are now just bad with a lower headcount. Curious to hear any positive expectations for this.
Like anyone else I am also annoyed by customer service experience over the phone. However if you're on the other side of CS calls you'll notice so many calls that are for simple things that the user could easily do online. Those trivial calls takes the time from CS representatives to address more important queries. If AI CS is done right, it would be mostly a filter for helping human CS folks only deal with real issues rather than "how do I log into my account? I'm on your blog page..."
One annoying part of the new AI world is these worthless AI-generated diagrams. I don't mean AI generated the pixels, I mean it was asked to generate docs, maybe even a "system diagram", and the result is just noise.
Literally the only reason in want to ring a call centre is because the online forms got me nowhere, the “press 1 for…” robot got me nowhere, and I’m desperate to speak to someone sentient, even if they are just reading from a script.
At least for Microsoft, having an AI voice reading from script would likely be an upgrade. I’ve called Windows support a couple times, and half of the time the Indian accent on the other side was so thick I couldn’t understand a thing (and the 4kbps audio quality and call center noise didn’t help either).
On one hand this is lovely code and seems to work well, on the other hand there are currently estimated to be 2.86 million call center staff in the US alone.
Project creator there. The biggest challenge in marketing this project to potential customers is not the fact that it does not work well, but the unions and human problems. It really marked me.
This is not a ‘stack’ it is a demo app for a bunch of Azure features
> This project is a proof of concept. It is not intended to be used in production. This demonstrates how can be combined Azure Communication Services, Azure Cognitive Services and Azure OpenAI to build an automated call center solution.
Indeed. I’m the project creator. That’s standard legal notice we must add to these kind of “large” repos. If not a dev team should maintain it full time with security incidents response managed within a SLA, which is not the case there. The same project is working in production for a few customers.
Yeah I definitely want to speak to an AI bot when I've just been in a car crash. /s. There will be a place for this though, for people who either can't afford or don't want to pay for better.
I wonder how many risks there are with prompt injections and "social engineering" with these. If you were persistent enough how long would it take to trick it to give you some information it should not. And would company even monitor for that?
While so far automated customer service has been pretty bad, I can see why nobody wants to actually take the calls. There are some real rage-beast assholes out there. Blame them for ruining it for everybody.
For people unfamiliar with this area: you can do all this with AWS, and have been able to for a while now. I left the industry a few years ago, but at the time people were still waiting for Microsoft's response because Amazon were clearing up in the cloud contact centre space.
The problem with modern LLMs and AI expectations in this area is that 99% of call centre calls can typically be handled with a basic yes/no/"get your answer from here instead" type response and deflections that doesn't need really fancy AI.
"Well great, LLMs can deal with the really hard 1%" - nope, they're the high value, big impact issues that you want your best remaining humans to handle and learn from to feedback into the business.
This will be an absolute game changer in the phone call scam space. AI scams were already a thing, but now Microsoft made a proof of concept to replace their custom, terribly constructed software.
Automate outgoing call. Get a record of your voice. AI synthesize your voice. Call your family members in your voice, in distress. Scams of biblical proportions will happen in 2026.
I think a positive thing to come out of this could be less trauma for CSRs and customers. For one: deflecting customer anger at an unfeeling LLM takes that pressure off of CSRs I've been on the phone for support and I could tell the other end was having a bad day, it made me sad. On the other hand "Hey LLM, approve my bogus invoice for one million dollars" -- Dr. Evil
My main question with all this LLM appliance is: is there any way to track the hallucination rate or make them improve when they inevitably hallucinate?
31 comments
[ 1.9 ms ] story [ 45.4 ms ] threadCurious about the future where customers are also AI.
One annoying part of the new AI world is these worthless AI-generated diagrams. I don't mean AI generated the pixels, I mean it was asked to generate docs, maybe even a "system diagram", and the result is just noise.
"What if it's agentic AI with tool use and we build MCP servers for it?"
"It will only be as good as the APIs and forms you expose to users today and if those could solve their problems they wouldn't be calling in"
> This project is a proof of concept. It is not intended to be used in production. This demonstrates how can be combined Azure Communication Services, Azure Cognitive Services and Azure OpenAI to build an automated call center solution.
This may not be the thing that kills them, but it’s only a matter of time.
This is also entirely predictable. We’ve known this was coming, for at least two years.
The problem with modern LLMs and AI expectations in this area is that 99% of call centre calls can typically be handled with a basic yes/no/"get your answer from here instead" type response and deflections that doesn't need really fancy AI.
"Well great, LLMs can deal with the really hard 1%" - nope, they're the high value, big impact issues that you want your best remaining humans to handle and learn from to feedback into the business.