The Layoff That Looked Right on Paper

**Stacie Baird** (0:07)
When something goes wrong in your business, really wrong, not like routine, run-of-the-mill, Monday morning wrong, really wrong, how do you get to the right person fast?
Welcome to The HX Podcast, this is Stacie Baird, and welcome to our series talking about the intersection of AI and human beings, technology and the human experience, and how we can navigate the waters that are unknown to all of us. I open with that question because it's one of the things that I keep thinking about lately. Because if we think about when stuff goes really left, how do we get to the right person?
It often, the answer is someone notices and says something.
So, I want you to sit with that for a second because that's not a system. That's a human being. And in the last two years, a lot of companies quietly automated that person out of existence without completely realizing what the scope of their role was doing to impact the organization.
This is episode two of The AI Head Count Trap, and today we're talking about escalation and instinct, and what happens when that escalation instinct disappears from your organization because you're automating process. Not bad to automate process, but it requires that you actually document thoroughly what occurs during the front end of that process.
AI is genuinely excellent at answering these tickets that come in, for example, resetting a password, processing a return, answering, where's my order? That work is largely solved, so I'm not going to spend any time arguing the pros and cons of automating that or using AI agents or doing anything there because this is my belief, like sometimes there are some things that people want fast, and technology is exceptional for that.
But there's a second skill buried inside frontline roles that almost never gets discussed because it's not really a task. It's discernment, it's a read, it's the ability to look at something and think, that's weird, that's different, that's not routine, that's a customer about to walk away from us. That's not a normal error message, that's the first sign of something bigger is broken, upstream, downstream, or sideways. That's escalation instinct, and here's the uncomfortable truth, it's built through repetition, through pattern exposure, through a person handling thousands of routine cases and slowly developing a nose for the ones that aren't routine. When you automate the routine cases, which again, do it, that's a good use of AI, you also remove the volume of exposure that people use, humans use, to build the instinct in whoever was left. If you cut off the team, radically reduce them down way too far in the process, you don't just reduce your head count, positive upside, you remove a sensor that's been in existence.
This is about understanding how to ensure that that sensor still exists. This is not a hypothetical question, this is already happening all over the place.
I want to name two real public examples because, again, I'm not theory, I'm not building this case for a podcast. This has really played out already, and it's something that as we look at this intersection, we should be considering, especially CEOs that care about results, CHROs and Chief People Officers who are being tasked with the work of understanding what to downsize. So Klarna, the Swedish fintech, cut hundreds of customer service roles and told the market its AI chatbot could handle the workload of 700 human agents. Satisfaction scores dropped and the CEO eventually acknowledged publicly that the company had prioritized cost over experience and started rehiring humans. So swinging back to the center, right? Salesforce cut its support org from 9,000 down to 5,000. Mark Benioff said outright in an interview that AI now handles about half of customer conversations, meaning humans are still doing the other half. That's not a company where AI replaced the function. That's a company that cut in half and discovered the ceiling was about 50 percent, not 100 percent. It's not an antidote. Gartner is projecting by 2027, half the companies that cut customer service jobs specifically because of AI, will be rehiring for the same or similar role. That's not a correction that's going to be made on the fringe. This is going to be an expected outcome for half the market. So here's what it looked like inside a company you wouldn't recognize by name. So I'm not going to talk about the company name, never do here. But a company running a customer support for a subscription product moved almost all first-line interactions to AI. It made sense. The vast majority of tickets were password resets, billing questions, straightforward stuff like all the stuff that we use AI chatbots for the last probably three to five years. The team went, at this case, from 14 people down to four. The four were good, smart, experienced, but here's what changed. Instead of handling a broad mix of tickets daily, the boring ones and the occasional weird one, they were now seeing what the AI had already flagged as needs a human, which sounds efficient, except for it meant they'd lost a baseline. They weren't swimming in normal, so they'd lost the ability to instantly clock what then was abnormal. Three months in, a customer wrote in with a complaint that was published, and on the surface read like a standard billing dispute. The AI routed it as such. It sat in a queue for two days. It wasn't a billing dispute. It was the first sign of a data issue affecting a batch of accounts. The kind of thing that a year earlier would have gotten pulled out of the queue in an hour by someone who developed a gut sense for this phrase means something's actually wrong. By the time it reached the right person, it had already spread across double the number of accounts. Not a massive crisis, but a two-day head start that should have never happened. The same pattern as episode one. The AI didn't fail doing its job. Nobody was negligent. The sensor was gone, and nobody had noticed it was attached to the headcount that they trimmed, because nobody evaluated the impact of that. So I want to use another example, and I'm going to use this example very personally. It's not a client. It's an experience I had, and I'm not going to say who the insurance provider is because, again, retribution is rough. So I don't know about y'all, but sometimes I want an agent. Sometimes I just want a quick answer. But I had an issue where I received an automated letter. I had received a text message. And man, y'all, I could not get to a human being about an issue resolution related to a child that no longer lives in my household, that no longer has access to my vehicles, that has his own residence, his own vehicles, his own. But he was still, even though I removed him from my insurance once, he was showing back up because it was automatically re-added to my policy on the renewal. Had I not caught it, I would have continued to pay that because I don't always look at the details. And I did pay it for some period of time, by the way.