UK companies are increasingly trying to present ordinary automation as artificial intelligence, according to PR professionals quoted by the Guardian. The habit has a neat label: AI washing.
The phrase means stretching, polishing or exaggerating an AI claim so a product sounds more advanced than it really is. Sometimes there may be a genuine bit of machine learning somewhere in the system. Sometimes the tool may simply be normal automation with a fashionable badge stuck on the front.
For ordinary readers, this matters because “AI-powered” is now appearing on products, services, apps, workplace tools and investment pitches. If the label is vague, it can make something sound cleverer, safer or more useful than it actually is.
What AI washing looks like
The Guardian reports that communications workers are being asked to pitch companies as AI specialists even when the link is thin. Examples in the article include familiar technologies being rebranded as AI, products being renamed around AI, and businesses trying to attach themselves to the broader AI boom because it attracts attention.
That does not mean every “AI-powered” claim is false. Plenty of tools now use machine learning, large language models, computer vision or prediction systems in meaningful ways. The problem is that the same language is also used for simpler software that follows rules, fills templates, scans documents or automates a narrow task.
The difference matters. A genuinely useful AI feature might help compare information, recognise patterns, draft a first version, summarise material or make a tool easier to use. A weak AI claim might simply describe something software has done for years, but with shinier wording.
Why companies do it
AI is fashionable, and fashionable words sell. They can help a company get press coverage, attract investors, reassure customers that a product is modern, or make a small update sound like a major leap.
There is also a defensive motive. If competitors are saying AI, a business may worry that not saying AI makes it look old-fashioned. That creates a loop where more products get described as AI even when the practical benefit is unclear.
This is not unique to AI. Businesses have done similar things with “cloud”, “blockchain”, “smart”, “green” and “digital transformation”. The difference is that AI can affect trust more directly because people may assume the tool is making intelligent decisions, not just running a process.
The simple questions to ask
The best response is not to reject every AI claim. It is to ask better questions.
What does the AI actually do? Does it generate text, classify images, predict demand, summarise calls, recommend products, detect fraud, automate admin, or something else? If the company cannot explain the feature in plain English, that is a warning sign.
What happens without the AI? If the product would work almost exactly the same without the AI label, the claim may be more marketing than substance.
Who checks the output? A tool that drafts, predicts or recommends still needs oversight. This is especially important where the result affects money, work, health, safety, housing, education or legal rights.
What data does it use? If a tool needs your documents, emails, recordings, shopping habits, location or customer records, the AI claim should come with clear information about privacy, storage and access.
Automation is not automatically bad
One trap is thinking that “just automation” means useless. It does not. A well-designed automated tool can save time, reduce repetitive admin and make a service easier to use.
The issue is honesty. If a product is a scanner, rules engine, scheduling tool or template system, it can still be useful. But calling it AI may lead users to expect flexibility, judgement or intelligence that is not really there.
That gap between expectation and reality is where people get disappointed. It is also where businesses can make poor buying decisions, especially if they pay extra for an AI-branded tool that does not solve a real problem.
How this affects work
Workplace AI claims deserve particular care. Staff may be told a new system is intelligent, objective or transformative, when in practice it may be a limited tool with blind spots. Managers may also use AI language to justify job cuts, restructuring or extra monitoring.
ManyHands has already looked at what UK workers should ask when a company blames AI for job cuts. The same scepticism applies to product claims. Ask what the system changes, what evidence supports it, what humans still decide, and what happens when it is wrong.
There is a useful parallel with checking AI tools before giving them access. The more a tool can see, change or decide, the more precise the explanation should be.
What good AI claims sound like
A stronger AI claim is usually specific. It might say the tool transcribes meetings locally on your device, detects duplicate invoices, suggests likely customer-service replies for staff to review, or summarises long documents while linking back to the source material.
That kind of claim gives you something to test. You can ask how accurate it is, what data it uses, whether it works in the UK, what it costs, what limitations are known, and whether a human reviews the result.
A weaker claim leans on broad language: AI-powered, intelligent, next-generation, revolutionary, autonomous, agentic, transformative. Those words are not proof. They are wrappers. The useful part is underneath, if there is one.
The takeaway
AI washing does not mean AI is fake. It means the label is being used too casually. That makes it harder for ordinary people to spot the tools that are genuinely useful.
The practical habit is simple: translate the marketing back into a job. What job does the AI do, with what data, under whose control, and with what evidence that it works?
If a company can answer that clearly, the AI may be worth a look. If it cannot, you may just be looking at old software in a shiny new hat.
Sources: The Guardian and UK government AI regulation white paper.
