The hardware behind the future


Layoffs Cite AI While Some Employers Rehire: Executives, Analysts and Lawyers Say Where the Line Sits

Layoffs now cite AI while some employers are rehiring. Analysts, founders, CEOs and lawyers told Autonomy Magazine where they say AI stops and human judgment begins.

In May, Cloudflare cut about 1,100 jobs, roughly a fifth of its staff, and its CEO said the vast majority of those let go were “measurers,” his term for middle management and for finance, legal and audit roles. A month later, Ford went the other way. After replacing staff with AI, it rehired 350 veteran engineers, and a company vice president said Ford had “mistakenly” thought that simply introducing AI would produce a high-quality product. Two moves a month apart show where the workplace argument stands: AI is changing which work people do, and employers are still learning where the line sits.

The cuts are real and they are large. Oracle disclosed in June that it had 21,000 fewer employees over twelve months, and its filing says AI adoption may continue to result in workforce reductions. Salesforce cut about 4,000 support roles after its AI agents reduced support cases, and IBM replaced about 200 HR roles with AI agents. The Financial Times counts nearly 140,000 U.S. tech job cuts so far this year. Some companies, Microsoft and Cisco among them, say their cuts are about moving money toward AI and not about AI doing the work. For the people laid off, the difference is small.

The research points to tasks rather than whole jobs. A study from Anthropic, the maker of Claude, measures how much of each occupation’s work AI is actually being used for. Computer programmers rank highest at 75 percent coverage, followed by customer service representatives, with data entry keyers at 67 percent. About 30 percent of workers have no measured coverage at all, including cooks, bartenders and lifeguards. The authors found no systematic rise in unemployment among highly exposed workers since late 2022, though they saw suggestive signs of slower hiring for workers aged 22 to 25, with the job-finding rate in exposed occupations down about 14 percent from 2022. They call the evidence limited.

Early reversals show the limits of replacement. Klarna began rehiring human customer service agents after its CEO said cost had been too dominant a factor and quality suffered. Commonwealth Bank of Australia restored 45 customer service roles it had given to an AI voice bot and called the move an error. IBM said it would triple entry-level hiring this year. The pattern is consistent: the best results come when AI takes the repetitive part of a job and a person keeps the judgment.

So where do employers say the line sits? We asked.

Autonomy Magazine put the same questions to company founders and CEOs, industry analysts, lawyers and other working professionals. Can AI replace the workforce, or should companies cut or not cut jobs because of AI? Can it enhance day-to-day work and data collection? And have they tested AI themselves, and would anything change their minds? What came back was a wide range of sectors and a surprisingly consistent answer. Their words follow in full.

We start with the analysts. Beatriz Valle, senior analyst for enterprise technology and services at GlobalData, said the layoff picture is harder to map than the headlines suggest, because cuts and AI-driven hiring are happening in the same places:

“AI-driven layoffs are concentrated in white-collar, digital, repetitive and cost-sensitive functions, while AI-driven augmentation is strongest where AI helps scarce or skilled workers increase output rather than replace them outright. Mapping the impact of AI on job losses can be complex because, despite the massive cuts seen in sectors like tech, AI is often also helping redesign functions and creating demand for AI-related skills. For example, IT operations and technology and media are some of the most disrupted sectors, but there is also AI-driven augmentation taking place rather than just cuts. In customer service and contact centers, although there have been cuts, the sector is more often redesigning work than eliminating roles. In data analytics, workers have suffered significant job losses, but again there is also strong AI adoption, in turn creating vacancies. Occupations like software development have seen net job gains, via code testing and bug detection and resolution.

It is also interesting to see that some companies are rehiring workers because their ROI calculations were overoptimistic, and this is happening across different sectors. Human expertise is becoming more valuable in some areas because experienced employees provide judgment and accountability. For this very reason, we are observing graduate jobs being impacted, although cuts are taking place across the board.”

Among the executives who responded, Nicholas Evans, founder and CEO of Thinkers360, pointed to an overlooked group of casualties, recent graduates and MBAs facing a hiring freeze, and to what his company’s research says about trust:

“While tech layoffs dominate headlines, an overlooked casualty is the hiring freeze for recent graduates and MBAs. Companies are navigating where the line sits between jobs AI can replace and those it cannot, but many are missing a critical truth: even automated workflows require human critical thinking to keep them on track.

Furthermore, our 4th annual Thinkers360 AI Trust Index reveals that over 80% of users and providers remain concerned about AI trustworthiness. Yet, encouragingly, over 71% agree that the benefits ultimately outweigh the risks. As employers sort through restructuring, sometimes using AI as cover for broader corrections, they must realize that long-term success relies on blending automation with genuine human expertise. Getting this balance wrong is the difference between sustainable growth and operational failure.”

Janine Yorio, CEO of Community Digital Technologies, looked at the data center industry, where automation helps with monitoring and forecasting but the buildout still depends on people:

“AI can replace certain repetitive tasks, but replacing employees is a much more complicated proposition. In the data center industry, automation can improve things like monitoring operational efficiency and energy forecasting. However, the industry’s expansion also depends largely on human expertise in building goodwill with local communities, which is most definitely a job that cannot be replaced by AI. AI can analyze enormous amounts of information, but it cannot build trust with local residents concerned about how a data center will affect their neighborhood.”

In automotive retail, two leaders at CarEdge made a similar case about service and accountability. Ray Shefska, the company’s co-founder, drew on decades in the business:

“Having spent decades in automotive retail, I can tell you that selling a car involves far more than providing a price or answering a question. AI can handle routine inquiries, organize inventory information, compare financing scenarios and reduce paperwork, but it cannot automatically replace the experience required to resolve complicated customer situations. When dealerships focus too heavily on cutting labor costs, they risk sacrificing the service and transparency customers expect. I see AI as a tool that can make knowledgeable employees more effective, not a substitute for having knowledgeable employees in the first place. The real test is whether the customer receives better information and a better experience.”

Zach Shefska, CEO of CarEdge, offered a way to measure it:

“AI has enormous potential to change automotive retail, particularly in customer service, pricing analysis, vehicle research and routine administrative work. But replacing human employees simply because a chatbot can answer questions is a risky strategy. Buying a vehicle involves substantial financial decisions, complicated negotiations and situations where consumers need accurate information and accountability. AI can help employees analyze dealership quotes, identify questionable fees and respond to customers more efficiently, but an automated response is only valuable if it is correct. I believe companies should measure AI’s success by whether it improves customer outcomes, not simply by how many positions it allows them to eliminate.”

Nidhi Singhvi, CEO of Unvault, described what AI does and does not change in gold valuation and resale:

“In gold valuation and resale, AI can dramatically improve the speed and accessibility of information that once required substantial manual analysis. Technology can help organize jewelry characteristics, estimate intrinsic metal value, compare market information and give consumers a clearer starting point before selling their possessions. However, AI-generated estimates should not be confused with a complete physical appraisal or a guaranteed purchase offer. Human expertise remains important for verifying authenticity, evaluating unusual pieces and resolving discrepancies. I believe the most productive use of AI is to automate repetitive analysis while allowing people to focus on specialized judgment, customer education and decisions where accuracy and trust are essential.”

Aaron Strout, author of Wired for Purpose, who has spent more than two decades in technology, marketing and business leadership, said AI is already replacing some work and cautioned companies against cutting jobs just because they can. He also described how he has tested it:

“AI is already replacing some roles, mainly those built around rote, easily repeated tasks. It’s also replacing some creative work, although with mixed results, and entry-level coding is among the hardest-hit areas. Where AI will not replace people is where mistakes put human life at risk, such as pharma, automotive, and safety equipment. It also can’t replace people where the work demands empathy, complex judgment, and other deeply human skills.

I’d caution companies against cutting jobs just because AI makes it possible. Short-term savings are easy to measure, but the loss of trust, creativity, and institutional knowledge is not. I’m a firm believer that AI is best used in collaboration with people, providing scale and speed while humans provide judgment. AI can and will do much of the heavy lifting that humans used to do, but it’s most effective when a human takes the last mile.

Absolutely, and in many cases it already does. Police departments use it to monitor video feeds, analysts use it to gather and summarize data in minutes instead of days, and developers use it for basic coding. Marketing teams use it to produce derivative assets for creative campaigns. In each case the tool removes the repetitive work and frees people to spend more time on the parts that require judgment and creativity.

I’ve used AI intensively for the past three years, including tools like Writer, ChatGPT, Claude, Perplexity and Gemini, and I’ve learned a lot about where it excels and where it falls short. Having grown up in the world of technology, I know you can never say never. With every new model, I’m surprised by how quickly AI is improving at skills I thought were safe.

Still, I believe networking, trust-building, navigating nuanced situations, and any job that requires empathy will remain human. What would change my mind is seeing AI show sustained, reliable judgment in high-stakes, ambiguous situations, with real accountability attached. I haven’t seen that yet.”

Chris Gillett, a headshot photographer and expression coach who runs Chris Gillett Studios, is already a user of AI and also a business it is challenging. He said the human part of his work is what clients are paying for:

“AI-generated photography is already challenging the traditional headshot business. But what it produces is a likeness, not a person. An AI headshot can look polished and still say nothing true about who you are, and people notice when the face on a LinkedIn profile doesn’t match the one that walks into the meeting. My job is less about the camera and more about the person in front of it: coaching expression, reading subtle shifts in body language, and helping clients come across as confident, likable and credible. AI can’t do that, because it never actually meets the client. That said, I’m not anti-AI. I use it every day. I edit video with AI tools, and AI assistants handle much of the administrative side of running Chris Gillett Studios, from scheduling to invoicing to email. That gives me more hours behind the camera, where the real work happens. What would change my mind? An AI that can make a nervous person laugh and actually relax. Until then, the human part of this job is the job.”

Two lawyers said the same thing about their own field: AI can speed up the paperwork, but it cannot take over the judgment. Russell “Rusty” Reynolds, JD, founding partner of the Reynolds & Reynolds Law Firm, spoke from personal injury practice:

“AI can help personal injury law firms organize medical records, summarize documents, identify relevant information and reduce repetitive administrative work. However, evaluating the full impact of a catastrophic injury requires understanding medical evidence, individual circumstances and the human consequences of an accident. The current state of AI is like knowing enough to be dangerous in part because it sounds authoritative but lacks common sense or situational awareness. Technology cannot independently replace the judgment involved in assessing liability, developing legal strategy or advocating for an injured client. My concern is that employers may mistake faster document processing for an ability to replace experienced professionals. AI should help legal teams devote more time to clients and complex cases, with appropriate safeguards for accuracy, confidentiality and accountability.”

Don Ford III, managing partner of Ford + Bergner LLP, spoke to estate planning and warned about what happens when positions are cut before their responsibilities are understood:

“AI can streamline legal research, document review, administrative work and the organization of complex financial records, but those efficiencies should not be mistaken for a replacement for legal professionals. Estate planning involves understanding family dynamics, anticipating disputes, interpreting individual circumstances and exercising judgment that cannot responsibly be delegated entirely to software. AI-generated documents may appear comprehensive while overlooking issues with serious financial and legal consequences. I believe law firms should evaluate AI based on whether it improves accuracy, efficiency and client service, while maintaining human oversight and professional accountability. Eliminating positions without understanding those responsibilities could create more problems than it solves.”

Across the sectors, the answers line up. The people we asked described AI as strongest on repetitive work such as records, scheduling, monitoring, summaries and first drafts, and weakest where trust, empathy, judgment and accountability are involved. Several warned against cutting jobs simply because AI makes it possible, and one noted that companies sometimes use AI as cover for broader corrections, which is the gap Ford, Klarna and Commonwealth Bank of Australia have already run into. The line, in other words, is not between jobs AI can touch and jobs it cannot. It runs through the middle of almost every job, between the repetitive part and the part that needs a person.

Leave a Reply