AI companies are making significant claims about automation replacing human workers, while evidence suggests the reality is more nuanced. Data visualizations reveal what's actually happening in the workplace as AI adoption accelerates across industries.
Artificial Intelligence (AI) companies are making vast claims about the ability of their tools to replace human labour.
Some jobs will be automated, others will be "augmented". The bosses of the world's biggest companies are diverting vast sums into these tools, partly with the knowledge that they could save money on headcount.
"Flat is the new up", we are told, in terms of the size of a company's workforce as investors ask whether jobs should be done by new recruits - or, instead, armies of "AI Agents", virtual workers tasked with doing specific roles, some of them relatively skilled.
If even half true there will be an impact on us all, across sectors and individual careers, and perhaps it will happen sooner than we think.
Nobel prize-winning economists recently warned the world “must act now”, external to ensure that AI leads to rising living standards and not large-scale job displacement, and last month London businesses warned they were struggling to find the skills they need as AI disrupts the jobs market.
It is early days yet, but there are already some notable patterns in the data.
This chart is the industry benchmark for how various models can perform the tasks previously done by humans, in this case using and developing computer software.
Full story reconstructed from BBC News. Formatting and media may differ from the original.
This measure showed that three years ago, large language models (LLMs) were only able to reliably complete tasks humans took seconds or minutes to do. Now they are increasingly able to complete fairly complex tasks taking an hour or so.
Now, some of the LLMs can find problems in a cryptocurrency contract and even develop and streamline the model itself, which would take a human several hours.
The latest generation of models could start to entirely develop themselves in the next year or so. This is just software coding, but the same type of pattern is being seen, at an earlier stage, with financial analysis, early stage legal work, even some entry level creative industry jobs.
What does that mean for jobs? The most thorough analyses out there come from the United States, using four years of data on employment outcomes by age among a range of occupations most exposed to AI (including software developers and customer contact reps) - and least exposed to AI (health workers, childcare workers, hairdressers).
Stanford University's analysis of wage and jobs data finds a hit to employment for 22 to 25-year-olds of 2.7% since ChatGPT became widespread, rising to 12.8% in the most AI-exposed sectors such as finance, software and creative industries.
Not all economists agree, arguing that other factors such as interest rate rises can explain this.
Online jobs postings have also been affected since the launch of ChatGPT and other LLMs.
There are many other factors here, from interest rates to taxes, however there is something of note in the OECD's recent analysis of the difference in jobs posting between its slightly different measure of highly exposed sectors (eg telemarketing and legal services) and less exposed sectors (construction, cleaning and food prep).
The UK was hit quite notably on this measure at a time when rates were stable or being cut. It also predates last year's National Insurance rise. On most international measures, the UK's service sector concentration leaves it exposed to potential AI jobs losses.
AI usage is measured in "tokens", which are small chunks of text that AI systems use to understand, analyse and generate language. On average, one token is roughly equivalent to three-quarters of an English word.
There have been astonishing increases in AI use in 2026, and the increase in token use has vastly outweighed the declining “per token cost”.
The world's top companies that are using AI had deployed "token leaderboards" to try to get their employees to create as much gain in productivity as possible from the most advanced models.
Trillions, and sometimes thousands of trillions of tokens (quadrillions) have been used over the past few months primarily for “agentic use”, which is for agents that can perform tasks automatically. The problem here is that incredible bills were racked up, so much so that many of those companies have now begun to ration the use of these models.
This is important. It might show that there are limits to how much work could be automated. The virtual workers might be more expensive than the human ones. It depends on the task.
One factor to watch out for however, is that many companies, including Western ones, are diverting to much cheaper forms of AI, derived from Chinese models that are provided freely on to the market.
So there is much uncertainty here, but some clear trends emerging.
