Major Australian tech company Atlassian has terminated 150 staff with a pre-recorded video, as the company pushes its AI-embedded customer contact solutions.
The stupid thing about replacing staff with “AI customer contact” is that the AI can only really spit out the same stuff you put in its knowledge base, i.e. the stuff that in the documentation in the first place (and maybe perform limited actions that also would have had to be implemented as forms). All it does is save someone broadly 0 seconds on what it would have taken to do a regular search of the documentation.
If I’m actually phoning up Atlassian to ask something it’s because what I want isn’t available online, and AI doesn’t solve that at all.
It’s the same thing with Stack Overflow. Usage in dropping because devs are relying on AI…. that got its information from SO. What’s gonna happen when you start running into new problems?
So, I agree with you, and I am the same way. But you and me, we represent like a fifth of support callers. AI could deflect an alarming number of daily support cases. Just finding information in the documentation often requires a deep and thorough understanding of the product, and it’s really difficult in documentation to separate “this is a common problem everyone has” from “this weird thing has never happened before and you need to talk to the dev who coded the fucker.” AI is fairly good at that level of pattern recognition.
The problem is that you still need the people to take the hand off, and deflection doesn’t mean they got the right answer, it just means they left.
[glancing around my office at all the people using jira] Yeah, sure. That’s the intention.
Seriously, though, I’m an “educated professional” with a liberal arts degree who uses jira every single day. Being an “educated professional” doesn’t mean you have PM skills, or tech troubleshooting skills, or know how to search documentation for your problems. Educated professionals are a cross section of the larger population, and are more or less a representational sample of the whole of humanity. There are proportionally as many people whodon’t know what a Kanban board is, or can’t figure out why they don’t have permission to delete the 350 epics they accidentally created.
AI assistance is like an interactive FAQ. It can do a little more than a static list of questions and answers, but the answers should also be validated by a human with the knowledge and understanding of the underlying systems. AI agents hallucinate and make up answers all the time. LLMs are essentially pattern recognition, and novel problems often break patterns. A human would go “huh, that’s weird.” An AI will classify a platypus as a duck and tell you with confidence how to pluck it.
Yeah but that’s nothing new, companies have paid millions in research to make the perfect telephony menu tree to force people to hang up or get misdirected instead of actually talking to a person. This is a legit thing that several companies offer services over
The stupid thing about replacing staff with “AI customer contact” is that the AI can only really spit out the same stuff you put in its knowledge base, i.e. the stuff that in the documentation in the first place (and maybe perform limited actions that also would have had to be implemented as forms). All it does is save someone broadly 0 seconds on what it would have taken to do a regular search of the documentation.
If I’m actually phoning up Atlassian to ask something it’s because what I want isn’t available online, and AI doesn’t solve that at all.
Humans are good at exception handling.
LLM isn’t. At all.
It’s the same thing with Stack Overflow. Usage in dropping because devs are relying on AI…. that got its information from SO. What’s gonna happen when you start running into new problems?
So, I agree with you, and I am the same way. But you and me, we represent like a fifth of support callers. AI could deflect an alarming number of daily support cases. Just finding information in the documentation often requires a deep and thorough understanding of the product, and it’s really difficult in documentation to separate “this is a common problem everyone has” from “this weird thing has never happened before and you need to talk to the dev who coded the fucker.” AI is fairly good at that level of pattern recognition.
The problem is that you still need the people to take the hand off, and deflection doesn’t mean they got the right answer, it just means they left.
Support callers for banks and ecommerce maybe, but Jira is a tool for educated professionals.
[glancing around my office at all the people using jira] Yeah, sure. That’s the intention.
Seriously, though, I’m an “educated professional” with a liberal arts degree who uses jira every single day. Being an “educated professional” doesn’t mean you have PM skills, or tech troubleshooting skills, or know how to search documentation for your problems. Educated professionals are a cross section of the larger population, and are more or less a representational sample of the whole of humanity. There are proportionally as many people whodon’t know what a Kanban board is, or can’t figure out why they don’t have permission to delete the 350 epics they accidentally created.
AI assistance is like an interactive FAQ. It can do a little more than a static list of questions and answers, but the answers should also be validated by a human with the knowledge and understanding of the underlying systems. AI agents hallucinate and make up answers all the time. LLMs are essentially pattern recognition, and novel problems often break patterns. A human would go “huh, that’s weird.” An AI will classify a platypus as a duck and tell you with confidence how to pluck it.
That’s why there are 3 tiers of support
I guarantee they aren’t replacing tier 3 yet, meaning that problems the AI can’t solve will theoretically see a human
but its also designed to discourage people from seeing a human, by adding so many layers to the AI CS that you give up anyway.
Yeah but that’s nothing new, companies have paid millions in research to make the perfect telephony menu tree to force people to hang up or get misdirected instead of actually talking to a person. This is a legit thing that several companies offer services over
You don’t need AI for that