 |
| - Image by Copilot |
The Promise
Technocrats and AI’s proponents tell us that the future looks bright – that AI will relieve us of tedious jobs, that it will cure cancer, stop world hunger and solve the world’s problems. Nor am I against those technologies. Assuming that the environmental issues have been resolved, there is nothing I would like better than having AI and robots taking over a lot of the routine, dangerous, soul -destroying or physically demanding work that takes such a toll on human beings.
The aim of this post then, is not to suggest that AI or automation are inherently negative or inevitable, but to look at the current trends critically and to proceed with caution so that society itself is not damaged in the rush to implement them. Part I will be about the way workplaces are changing and Part II will be about ways to mitigate its worst effects.
The Reality
There are workplaces which seem to embody the dream. I am thinking for example of the state -of -the -art Aged Care facility in Aalborg, Denmark, where high -tech sensors, AI and robots are being deployed to enable greater independence and less supervision, leaving the staff to provide the “warm hands” care such as physio, nursing and occupational therapy.
It is the result of a collaboration between the city, the university and other educational institutions which ensures that training and workplace fit occur smoothly, swiftly and efficiently. However, that is not the way things are playing out in most of the world.
Logistics
My second example was going to be German logistics and delivery company DHL because its website shows smiling humans directing AI -enhanced robots doing the heavy lifting – the unloading, storing and retrieving as well as analysing storage configurations for greater efficiency. On further investigation, however, I found out that in 2024, DHL sacked around 8,000 workers, ostensibly because of declining letter rates, but it had also added 8,000 robots. This seems to be the more general pattern. Once again, it is not so much the process, but the scale and speed of change.
Although there is some discussion about retraining employees, detail on how many workers have been successfully retained is scant. The company has also expanded its robotics investments, including a $737 million investment in UK robotics expansion, and expects to be using 1,000 AI-powered warehouse robots by 2030.
DHL is not alone. Other logistics companies such as Amazon soon followed, announcing in March 2024, that it would be cutting 16,000 jobs and spending $100 billion on AI. In November, 2025, United Parcel Service (UPS) announced 48,000 job cuts in logistics, including the 1,400 operations jobs announced in May 2025. In the same month FedEx announced that 2,000 positions would be cut. Watch this small clip of the Ocado warehouse in the UK and see if you can spot how many humans are needed to move over a million items a day. A rough tally of layoffs in the sector, suggests that more than 500,000 jobs have been lost in logistics alone in the last few years.
Given the conditions in some of those warehouses, some would argue that those jobs aren't a great loss. In the United States, for example, Amazon warehouse workers reported having to wear adult diapers to meet the company's productivity targets, especially with bathrooms often located far from their work stations.
In India, during a 2024 heatwave with temperatures reaching 50°C, Amazon workers were required to pledge they would take no breaks — not even to drink water — until they had met their daily targets. The company allowed only 5% downtime in a ten-hour shift.
However, that argument offers little comfort to displaced workers, their families, and the communities which depend on their income, especially in places where such a facilities were the major or only employer. It also ignores a harder economic truth. Desperate workers competing for fewer jobs is precisely what keeps wages low and increases that gap between the haves and the have nots which has been becoming increasingly apparent over the last couple of decades. Nor was logistics the only sector affected.
Other Sectors
Soon automation and AI were being applied to any number of other routine tasks such as data entry, customer service and banking, causing major job losses in other industries. Since 2022, the widespread adoption of generative AI which can produce images, code and audio in response to a prompt, via services such as ChatGPT, led to more layoffs. Tech companies, Telecommunications, Media and Entertainment, Financial Services and Banking were all hard hit. Dell for example, announced the layoff of 11,000 workers in March 2024 and Microsoft 8,750. Intel followed in May 2024 with 15,000, to focus on AI.
At Meta roughly 33,000 positions have been eliminated since 2022 — 11,000 in late 2022, 10,000 in 2023, approximately 3,600 in early 2025, and 8,000 in the April/May 2026 round. The smaller 2026 rounds (Reality Labs ~1,700) bringing the combined total to around 35,000.
Alphabet, Google’s parent company, began cutting 12,000 jobs in January 2023— around 6% of its global workforce — citing economic challenges and the need to streamline operations. Since then the cuts have been smaller and rolling rather than announced in big rounds — approximately 1,500–3,000 affected across advertising sales, cloud, cybersecurity and HR functions in 2025–2026, with voluntary exit packages also offered. Google is notable by its relative absence from the mass layoff lists — they've done it quietly and in smaller tranches rather than headline announcements.
Amazon confirmed a further 16,000 more corporate job cuts in January 2026, which combined with October 2025's 14,000 and the 27,000 positions the company eliminated in 2022 -2023 brings its combined total to approximately 57,000–75,000, depending on whether you include warehouse/operational and contractor roles.
Not all layoffs were specifically because of AI. Some of those in the postal services were because of falling demand for letter delivery. Massive job losses in the Automotive industry for example, were largely attributed to competition from China and falling sales. Airlines were still struggling in the wake of Covid, while those in clothing, retail and make -up (Estée Lauder for example shed 7, 000 jobs) were very likely secondary casualties due to a downturn in demand.
Less money in circulation means less discretionary spending, putting off new
car purchases, cutting out that morning latte and so on. Others were the result
of mergers and acquisitions or simply to make their balance sheets look better. DHL's share price jumped as soon as it announced its big layoffs.
Some more
cynical commentators claim that AI was the scapegoat for layoffs that would have been done anyway and yet others say, that
workers were pre-emptively laid off, so companies could afford to replace
workers with AI. By May 2025 several commentators noted that entry -level jobs were disappearing and new graduates were having difficulty finding jobs.
No one was particularly alarmed at this because it was thought that “knowledge workers” and professionals would be safe. As recently as January 2025 the World Economic Forum’s “The Future of Work Report” was still predicting that although 92 million roles would disappear by 2030, 170 million new jobs would be created — a net gain, on paper, of 78 million jobs and was urging people to upskill and become prompt engineers - the people who tell AIs what to do. However, beneath the surface AI was already changing.
The Coming of Agentic AI
Until around mid-2025, the Large Language Models (LLMs) which underlie AI, could only solve low level and routine problems and only after lengthy training, and still had to be prompted by humans. Now they were no longer simply reactive, but were becoming increasingly capable of autonomous reasoning across multiple steps – they could make plans, take action, access IT tools as needed, browse the web, write code and execute it, do their own error checks and produce a finished document or a conclusion, all without needing help from a human.
In 2024 the best AIs could only solve 2% of a set of 350 maths problems - called Frontier Maths, which were set by researchers and specifically chosen because they were unlikely to have appeared in any training data, but by January 2026 they could solve 40% of of these problems including post PhD level ones and some which had defied top level mathematicians. On the 20th of May, 2026 just over a month ago, Open AI solved the Distance Unit Problem, an 80 --year-old mathematical conundrum posed by Paul Erdős in 1946, using a "cross disciplinary combination of geometry and number theory that few human specialists possess"and obtaining a result which stunned researchers who thought it would not be achieved within their lifetime.
The Tsunami
By the end of 2026 an estimated 85 million jobs will have been displaced globally by AI and automation. The tech sector continued to bleed. In the first six months of 2025 alone, nearly 80,000 tech jobs were directly attributed to AI-driven layoffs. Meta for example, announced in March 2026 that it would be laying off 16,000 workers. IT consulting firm, Accenture announced it would be laying off 11,000 workers. Dell announced in March this year that it would be laying off 11,000, Cisco said it in May, that they would be culling 4,500 jobs and Oracle announced in June, that it would be axing 21,000 jobs.
Customer service roles were being eliminated in favour of AI-powered Chatbots as for example at Salesforce, a cloud -based software company which said it would be cutting 8,000 jobs with an additional 4,000 positions to follow in September 2025, because AI was now doing between 30 and 50% of its work.
Wall Street banks announced plans to shed approximately 200,000 jobs over the next three to five years, initially targeting entry-level and back-office roles first.
The tidal wave of disruption now spread into other sectors beyond those traditionally associated with automation. They included fields such as Architecture, Law, Engineering, Healthcare, Education, Retail and E-Commerce – Sainsbury’s UK, for example, announced that it was reducing its workforce by 20% or 3,000 workers, Insurance, Biotech, Pharmaceuticals and Education.
- International law firms such as Reed Smith and Allen & Overy have reduced junior and mid-level intake as AI handles document review and contract analysis.
- Accounting giants Deloitte and PwC report reductions of 10–15% in junior auditing and tax preparation roles in their UK and US offices since 2023–24
- Engineering and architecture firms including AECOM and Foster + Partners have cut drafting and design staff as AI-assisted tools absorb routine work
- Healthcare administrators, educators and advertising agencies tell the same story — United Health has reduced medical coding roles by 18%
- In Education, Pearson has cut administrative and tutoring staff
In the Music Industry -Michael Smith was charged in 2024 with having created hundreds of thousands of AI-generated songs and using bots to stream them billions of times, fraudulently obtaining more than $10 million in royalties that should have gone to real musicians.
In April 2025, streaming platform Deezer estimated that 18% of content uploaded to its platform every day is AI-generated — about 20,000 tracks daily. Since streaming platforms have a finite royalty pool, every fraudulent stream diverts income from living artists.
Counter Moves
Not everyone is taking these changes lying down. Robot law firm DoNotPay, which was offering free online legal advice for modest legal issues was taken to court after heavy lobbying by the legal profession claiming that, not having passed the bar exam or having a practising attorney on board, the company had no legal right to practice law.
While there have been some new jobs created in the tech sector, they are generally short -term, gig -style jobs with recent graduates teaching Large Language Models their newly acquired professional skills in particular fields. The work is insecure and highly competitive, but for the most part, the mass of new jobs has failed to materialise. Even companies which had no plans to sack workers, were now obliged to automate and use AI simply to remain competitive or to remain in business at all, though it also comes at great cost.
Some companies such as Deloitte have gone to great lengths to retrain their workforce, yet only 25% of those workers were subsequently re -employed.Alarm Bells
Even before Agentic AI was released, its founders and industry insiders were beginning to express concern about where things were heading. In March 2023 more than 1,000 technology leaders and researchers signed an open letter calling for a six-month moratorium on AI development, warning that the technologies pose "profound risks to society and humanity."
Geoffrey Hinton, who is regarded as one of the "Godfathers of Deep Learning" and shared the 2018 Turing Award (the Nobel Prize of computing) for laying the foundations of AI for his pioneering work on deep learning, neural networks and artificial intelligence, resigned from Google in May 2023, so that he could speak freely about the dangers of AI, saying he now regretted his life's work.
He voiced concerns about deliberate misuse by malicious actors, technological unemployment, and existential risk from artificial general intelligence. Hinton went even further — he said "My greatest fear is that, in the long run, it'll turn out that these digital beings we're creating are just a better form of intelligence than people. Specifically with respect to agentic AI Hinton remarked, "Once systems are making autonomous decisions and acting in the world, the question of whether humans remain in meaningful control becomes acute.”
Ilya Sutskever, an Israeli-Canadian, co-founder and former Chief Scientist of OpenAI and along with Hinton, left OpenAI in May 2024 following internal debates about the direction of AI development and alignment protocols. What followed was damning. The other leader of the safety team, Jan Leike, sharply criticised the company on his way out, saying "over the past years, safety culture and processes have taken a backseat to shiny products." Then OpenAI effectively dissolved the entire superalignment team — the group specifically tasked with ensuring AI safety, although parts of it were integrated into other research efforts.
Sutskever went of to found Safe Superintelligence Inc., with a singular focus — creating superintelligent AI that would not pose an existential threat to humanity. Its first product will be safe superintelligence, and it will do nothing else until then. Although it has shipped nothing, the company is currently valued at $32 billion. It is a deliberate statement of principle.
On April 7, 2026, Anthropic announced its latest model, Claude Mythos Preview, which it described as its most powerful yet — but withheld it from public release because of serious concerns about its ability to breach computer security. In tests, it found critical faults in every widely used operating system and web browser, 99 percent of which had not yet been patched. Instead, Anthropic formed Project Glasswing, partnering with Amazon, Apple, Google and Microsoft to use Mythos to identify and fix security vulnerabilities rather than release it broadly.
Other Reasons to Treat AI with Caution
Note how recent all this is. The field is moving very fast and both its scale and speed are rapidly outstripping the capacity of humans and governments to respond. It could be, however that AI itself is creating conditions for a bit of a slowdown.
· The rapid roll out of data centres has hit a bit of a roadblock as communities protest against excessive use of water, power and space or the infrastructure demands are simply so large that they exceed the capacity of electricity grids and especially renewable energy supplies to keep up. [Read the earlier post about this here].
· Some of the results and documents are not always as accurate as they could be. In my own usage I have frequently been given references which have proven incorrect. A friend of a friend recently received an AI X -ray diagnosis stating that she had a broken elbow, when she in fact had a broken arm, highlighting the need for human oversight. Good thing there was no surgery involved!
The landmark case in legal proceedings was Mata v. Avianca (2023) in which attorneys submitted a legal brief citing six court cases that did not exist, entirely fabricated by ChatGPT, complete with realistic-sounding case names, docket numbers, and judicial opinions. When confronted, one attorney even asked ChatGPT to confirm the cases were real — and the AI obligingly confirmed its own fabrications.
This has only escalated since. A federal judge in Oregon sanctioned two lawyers $110,000 — the largest AI hallucination penalty in American legal history — after they submitted 23 fabricated citations and eight invented quotations, and the case was subsequently dismissed. A researcher tracking such cases found more than 1,659 cases globally where a court or tribunal has commented on AI-generated hallucinations in legal filings.
In Education, Canadian provinces changed their AI policy, after finding that a flagship report recommending AI was written by an AI Chatbot and contained 15 false citations.
South Africa scrapped its national AI policy after discovering similar issues.
· Returns on investment (ROI) have not always lived up to expectations, and profits did not necessarily improve despite widespread layoffs. This is partly because serious AI adoption is more expensive and ongoing—it relies on subscription models rather than one-time purchases, as was typical with traditional software suites.According to a National Bureau of Economic Research (NBER) survey of 6,000 CEOs in four countries, conducted earlier this year, 90% saw no measurable improvement on either employment or productivity over the last 3 years.
· AI tools are only as effective as the operator using them and the source information available to it. Generic packages often fail to adapt seamlessly to every business, as success depends on customisation, expertise, and alignment with the business itself.
A Glimmer of Hope
The good news is that several companies such as Uber and Salesforce have begun to rehire some of their workers. It turns out that customer -facing work sometimes needs actual humans to manage customer service. However, these rehires are a drop in the ocean compared to the scale of layoffs already underway. Meta and IBM have also rehired some IT professionals for roles in AI and cloud computing. [Good luck trying to open any of these links too, though subscribers may be more successful].
These leaves the Question, "What do we do with all the displaced workers? " which we will look at in the next Post.
Let me repeat - I love AI and use it for everything from earnest fact checking and formatting to mundane things like how to get the black gunk off the iron, and can see enormous potential for good, but as it stands, without safeguards and human oversight, we get things like Australia’s disastrous Robodebt scheme, which used impenetrable algorithms to raise false debts against welfare recipients, causing many of those who did not have the means to pay or seek legal recourse, to take their own lives.
“Moving fast and breaking things” as Mark Zuckerberg said in his letter to investors in 2022, might be fine if you are just launching a shiny new gizmo, but less so when it comes to one which is likely to affect large numbers of people and the societies we live in.
No AIs were harmed in the writing of this post, nor as far as I can tell, were any humans. Without AI, this post could not have been written. The image is by Copilot, background research and references have mainly been done by Claude, also Ecosia and ChatGPT, especially when I have used up my quota. Proofreading has also been done by Claude, so while I would like to blame it for any errors, they are probably mine.
Comments