In February 2024, Klarna’s CEO Sebastian Siemiatkowski told the world that an OpenAI-powered chatbot had replaced the equivalent of 700 customer service agents. The company celebrated projected savings of $40 million. The fintech industry took notes.
Fourteen months later, Siemiatkowski told Bloomberg the opposite: Klarna had gone too far, quality had dropped, and the company was rehiring human agents.
Klarna isn’t an outlier. It’s the most visible example of a pattern now playing out across industries. Companies that moved aggressively to replace human workers with AI are quietly (and sometimes not so quietly) bringing people back.
The data tells the story at scale: 29% of companies that cut staff for AI have already rehired for those same positions, according to staffing firm Robert Half. Another 35.6% rehired more than half of those they’d fired. And one in three employers spent more on restaffing than they saved from the original layoffs. Separately, 55% of executives now say they regret their AI-driven layoffs, according to workforce analytics firm Orgvue.
The lesson isn’t that AI doesn’t work, but rather, AI without trained people doesn’t work. And every property management company evaluating how AI fits into their operation should be paying attention.
The Companies That Learned the Hard Way
The reversal isn’t confined to one industry or one type of role. It’s happening everywhere companies tried to substitute AI for human judgment.
Klarna is the case study everyone references. The Swedish fintech replaced roughly 700 customer service positions with an AI assistant built on OpenAI. The chatbot handled volume well, routine queries, password resets, basic account questions. But when customers had complex issues, disputes, or situations requiring empathy, the AI produced generic, repetitive responses that frustrated users and drove complaints up. By mid-2025, Klarna began rehiring human agents. Siemiatkowski’s admission was direct: the company focused too much on efficiency and cost, and the result was lower quality that wasn’t sustainable.
IBM deployed AI across its human resources function, and the system handled roughly 94% of routine HR requests. But the remaining 6% (the cases involving ethical judgment, sensitive employee situations, and nuanced decision-making) exposed the limits of automation. IBM subsequently announced plans to triple its US entry-level hiring across all business units in 2026. The company’s chief human resources officer framed it as a pipeline issue: without investing in human talent now, there would be no experienced professionals to draw from in three to five years.
Ford rehired 350 veteran engineers to work on vehicle quality problems that automated systems couldn’t resolve. The company’s VP of vehicle hardware engineering acknowledged that Ford had wrongly assumed feeding AI its design requirements would produce high-quality output without experienced human oversight. The quality issues that surfaced required the kind of pattern recognition and judgment that comes from decades of hands-on engineering experience, exactly the expertise that had been cut.
McDonald’s tested an AI-powered drive-through ordering system and shut it down after the technology repeatedly failed in ways that were immediately obvious to customers, including adding hundreds of dollars of chicken nuggets to simple orders. Human cashiers returned to the drive-through lanes.
These aren’t small experiments or pilot programs. These are flagship AI deployments at some of the world’s most visible companies, championed by their CEOs, executed across core business functions, and then reversed when the results came in.
Why AI Replacement Fails Where AI Augmentation Succeeds
The pattern across every reversal is consistent: AI handled the routine volume well and broke down on the complex, judgment-dependent work.
This isn’t surprising if you understand what AI actually does versus what it appears to do. AI is extraordinarily good at processing volume: triaging inbound requests, matching patterns, classifying data, and generating responses based on training data. It can do these things faster and more consistently than any human team.
What AI can’t do is read context that isn’t in the data. It can’t recognize that a long-time customer’s frustration isn’t about the specific issue they’re reporting, it’s about a pattern of unresolved problems that has eroded their trust. It can’t adjust its approach mid-conversation when the emotional temperature of an interaction shifts. It can’t exercise the kind of ethical and relational judgment that the IBM HR team needed for that critical 6% of cases.
Forrester Research predicted that roughly half of all AI-attributed layoffs would be quietly reversed. That prediction is playing out in real time. The companies now performing best aren’t the ones that replaced the most people with AI. They’re the ones that deployed AI to handle volume and kept trained people on the work that requires judgment, empathy, and contextual understanding.
As Klarna’s CEO put it after the reversal: in a world where AI can handle basic service, human service becomes the premium, the thing that builds trust, retains customers, and differentiates one company from another.
What This Means for Property Management
Property management is a relationship business wrapped in operational complexity. The parallels to Klarna’s experience aren’t abstract, they’re structural.
Consider the PM functions where AI is already being deployed: inbound call handling, maintenance ticket triage, tenant communication, and basic inquiry resolution. AI handles the volume layer well. Calls get routed. Tickets get categorized. Routine questions get answered around the clock.
But property management has its own version of Klarna’s 6% problem, and it’s much larger than 6%.
The owner who calls about a discrepancy on their statement isn’t looking for a data lookup. They’re looking for confidence that someone competent is watching their asset.
The frustrated resident whose maintenance request was closed in the system but not actually resolved needs a person who can listen, acknowledge the failure, and coordinate the fix, not a chatbot that generates a new ticket.
The leasing prospect who’s hesitating needs a human who can read the hesitation and adjust the conversation. The vendor relationship that requires negotiation, accountability, and follow-up can’t be automated without losing the leverage that comes from a real person managing it.
These interactions are where owner retention, resident satisfaction, and operational reputation are built or lost. And they’re exactly the interactions that break when you replace people with AI instead of pairing them together.
The Model That Actually Works
The companies emerging strongest from the AI adoption wave aren’t the ones that automated the most aggressively. They’re the ones that figured out the right division of labor between AI and people.
The framework is straightforward:
AI handles the volume.
Inbound calls get triaged automatically. Maintenance requests get categorized by urgency and routed to the right queue. Routine tenant inquiries get resolved through AI-powered chat. Rent reminders go out on schedule. Data entry and transaction classification happen in the background. This is where AI delivers genuine value, it processes more, faster, without fatigue, around the clock.
Trained people handle the judgment.
A remote accounting professional catches the coding error that would have distorted an owner statement, something the AI processed correctly by its own logic but got wrong in context. A leasing specialist reads a prospect’s hesitation and adjusts the conversation. A maintenance coordinator follows up personally with a resident after a repair to confirm the issue is actually resolved, not just closed in the system. A back office professional manages the vendor relationship that requires human accountability.
The combination scales.
AI gives the team capacity to manage more units without proportionally growing headcount. People give the operation the credibility, empathy, and judgment that retains owners and residents. Neither works as well alone as they do together.
This is the model that AppFolio’s 2026 Property Management Benchmark Report supports: firms broadly adopting AI expect 31% portfolio growth, and 34% of those same firms plan to increase headcount, not reduce it.
The most successful AI adopters aren’t cutting people. They’re redeploying them to higher-value work.
The Cost of Getting It Wrong
The Klarna story has a financial coda that’s relevant for any PM operator weighing how far to push automation.
One in three companies that laid off workers for AI spent more on restaffing than they saved from the original layoffs. The math is punishing: you lose institutional knowledge when people leave, you pay recruiting and onboarding costs to bring new people in, and you absorb the productivity gap while they get up to speed. For PM companies, where a single experienced accounting professional knows every owner’s reporting preferences, every property’s chart of accounts, and every vendor’s payment terms, the institutional knowledge loss is especially expensive.
Gartner predicted that by 2027, half of companies that cut customer service staff for AI would need to rehire. The smart move isn’t to wait for that prediction to come true in your own operation. It’s to build the right model from the start, one where AI amplifies human capability instead of attempting to replace it.
How Anequim Builds This Model
Anequim’s remote professionals are the human layer in the AI-augmented model, trained specifically for property management workflows, working inside your PM software (AppFolio, Buildium, Rent Manager, Yardi), and operating during US business hours in the same time zone as your team, your owners, and your residents.
The model includes AI-powered call center support for volume and after-hours coverage, paired with trained remote team members who bring the PM knowledge, adaptability, and relationship skills that Klarna, IBM, Ford, and McDonald’s all discovered AI can’t replicate.
That’s not a philosophical position. It’s what the data from the largest AI deployments in the world now supports: AI handles the volume, trained people handle the judgment, and the combination scales in ways that neither can alone.
Schedule a free strategy call to see how the model works for your portfolio.
