Will AI take my job?
That distinction between exposure and replacement is where nearly all the public conversation goes wrong, so it is worth being precise about it before anything else.
We build automation for UK businesses. We are the people brought in to remove manual work, and we have watched what actually happens on the other side of it. What follows is not reassurance and it is not alarm. It is what the credible research says, plus what we observe on live projects.
What the research actually says
Four sources carry real weight here. The figures are widely misquoted, so each one is stated as its authors stated it.
IMF, January 2024 — Gen-AI: Artificial Intelligence and the Future of Work, Staff Discussion Note SDN/2024/001. Around 40% of jobs globally are exposed to AI, with exposure concentrated in advanced economies because they have more of the cognitive, white-collar work AI touches. Critically, the IMF's own framing is that roughly half of exposed jobs see AI as a complement, raising productivity, and half face displacement pressure. The headline "40% of jobs at risk" is a misreading of this paper.
IPPR, March 2024 — Transformed by AI. Modelled a worst-case second-wave scenario in which up to 8 million UK jobs could be at risk in the absence of government action. This is a conditional scenario, deliberately constructed as an upper bound to argue for policy intervention. It is not a forecast, and anyone quoting it as one is misrepresenting it.
Goldman Sachs, March 2023 — estimated 300 million full-time jobs globally exposed to some degree of automation, with around two-thirds of US occupations exposed at least partially. Again: exposed, meaning some tasks within the role are automatable. Not eliminated.
Department for Education, November 2023 — The impact of AI on UK jobs and training. This is the best UK-specific occupational analysis available, and it is the source you should reach for on British data.
One correction worth making because the error is everywhere: the ONS does not publish AI job-exposure statistics. It publishes AI business adoption data, which is a different thing. Any article attributing a UK job-loss percentage to the ONS has invented the citation.
Exposure versus replacement: the distinction that matters
Jobs are bundles of tasks. AI does not take jobs, it takes tasks — and whether that costs you your role depends entirely on what is left in the bundle afterwards.
A conveyancing solicitor spends time reading documents, checking for anomalies, advising clients, negotiating and carrying professional liability. AI is very good at the first, useful on the second, poor at the third, bad at the fourth and legally incapable of the fifth. That role gets faster and shifts toward advice and judgement. It does not vanish.
A role whose bundle is entirely the first two tasks is in a different position.
So the useful question is not "will AI replace my job?" It is: what fraction of what I am paid for is task-shaped, and what fraction is judgement, relationship, accountability or physical presence? High task-fraction roles change most. Roles where a human is accountable for a consequential decision are the most durable, because accountability cannot be delegated to a statistical model — and under UK GDPR, significant automated decisions about individuals carry specific legal obligations that keep a person in the loop by design.
Which jobs are most and least exposed?
| Exposure | Characteristics | Examples | What actually happens |
|---|---|---|---|
| High | Text or data in, text or data out; standardised; digital-only; no accountability attached | Basic copywriting, data entry, first-line ticket triage, transcription, routine translation, template document assembly, entry-level research | Volume of work per person rises sharply; junior headcount growth stalls first |
| Moderate | Substantial task automation but judgement, client trust or liability remains | Software development, paralegal work, accounting, marketing, HR administration, graphic design, financial analysis | Role reshapes around review, judgement and client relationship; output expectations rise |
| Low | Physical presence, manual dexterity in unpredictable environments, regulated accountability, or the relationship is the product | Trades, healthcare delivery, skilled installation and repair, senior leadership, complex negotiation, teaching, social care | Largely unaffected directly; often gains from admin being automated around them |
The pattern in that table is the important part, not the individual examples. Exposure tracks how digital and how standardised the work is — not how skilled or how well paid. A senior copywriter is more exposed than a junior plumber, which inverts most people's intuition about what safe work looks like. If you run a business built on trades work, our automation for trades page covers where the admin time actually goes.
Will AI replace programmers?
No, but it has already changed what a programmer is paid for, and the change is not evenly distributed.
What AI does well: boilerplate, test scaffolding, translation between languages, explaining unfamiliar code, routine refactoring, first drafts of well-specified functions. On these, a competent developer with good tooling is dramatically faster than one without.
What it does badly: knowing what to build, system architecture under real constraints, debugging emergent behaviour across services, deciding which requirements are wrong, and owning the decision when a deployment takes production down at 2am.
We build automation systems for a living, with AI assistance throughout — including on the platforms we compare in Zapier vs n8n — and the bottleneck has never once been typing speed. It is knowing which of four architectures survives contact with the client's actual data, and which requirement in the brief is quietly incoherent. AI does not help much with either.
Where the real pressure sits is the entry level. The traditional route into software — junior does the simple, well-specified tasks while learning judgement — is precisely the work now most easily automated. That is a genuine structural problem for the profession, and it is not solved by individual developers getting better at prompting. It is a training-pipeline problem, and the industry has not addressed it.
How to make money using AI
Setting aside the online-course economy, three routes generate real income, in descending order of reliability.
1. Apply it inside expertise you already have. The reliable version. A recruiter who automates CV screening and reclaims eight hours a week is more valuable and can carry more clients — see how this plays out for recruitment agencies. An accountant who automates document chase does more advisory work at higher margin, as we cover for accountants. The AI is not the product — your expertise, delivered at greater volume, is. This works because you already have the two hard things: domain knowledge and clients.
2. Sell implementation to businesses that have neither the time nor the inclination. There is real, currently unmet demand from UK SMEs who know they should automate something and have no idea what. The barrier is not the technology — it is understanding operations well enough to diagnose the right problem. Which is why most of this market's failures are technically competent builds of the wrong thing.
3. Build products. Highest ceiling, lowest success rate, and the honest note: "AI wrapper on a prompt" is not defensible. If your entire product can be replicated by someone pasting your prompt into a free chatbot, it has no moat. The defensible versions own proprietary data, a distribution channel, or a workflow deeply enough embedded that switching is painful.
What consistently does not work: generic AI content at volume for its own sake, reselling access to tools people can use free, and being an intermediary who adds no judgement.
A checklist for the next eighteen months
Ordered by return on effort, not by how impressive it sounds.
What we actually see on projects
Three patterns, consistently, across the automation work we deliver.
Roles change more than they disappear. In most builds, the person whose manual work got automated ends up doing the higher-value version of the same function. The work removed is almost always the work they disliked — rekeying between systems, chasing documents, copying data.
The businesses that handle it badly are the ones that treat it as a headcount exercise. Automation as pure cost-cutting tends to remove the people who understood the process, leaving a system nobody can maintain and no one who can tell when it is producing nonsense.
The gains are rarely where anyone predicted. Time savings show up in unglamorous places — the eleven minutes per enquiry spent re-entering the same details into a third system. Nobody's forecast identifies that. An audit does.
The honest summary
AI is not coming for most jobs wholesale. It is coming for tasks, unevenly, and fastest for standardised digital work with no accountability attached.
The people who struggle will mostly not be those whose jobs were automated. They will be those whose roles were entirely task-shaped and who did not notice in time. The people who do well are those who understood their own process well enough to automate the boring parts and keep the judgement.
That is not a comfortable answer but it is an actionable one, and it is the same advice we give business owners: understand where the time actually goes before deciding what to change.
Start with the audit, not the software
If you run a business and the question is which roles to automate around rather than whether to, the sequence matters. Diagnose first, then build.
Thirty minutes on a call, via our free automation audit. We map how work moves through your business, show you where the time is going, and tell you what is worth building — and what is not. You keep the roadmap either way. If clinics and practices are more your world, our dental practices page shows what that looks like in that setting.
Thirty minutes, no obligation, and you will leave with a clear recommendation either way.
Will AI take my job? FAQ
Will AI take my job?
Probably not, but it will likely change it. The most-cited global estimate, from the IMF in January 2024, is that around 40% of jobs worldwide are exposed to AI. Exposure is not replacement — the IMF expects roughly half of exposed roles to be complemented rather than displaced.
Which jobs are most at risk from AI?
Jobs that are text or data in, text or data out, standardised, digital-only and carry no accountability — basic copywriting, data entry, first-line ticket triage, transcription, routine translation and template document assembly. Roles with physical presence, regulated accountability or where the relationship is the product are the least exposed.
Will AI replace programmers?
No, but it has already changed what a programmer is paid for. AI is good at boilerplate, test scaffolding and routine refactoring, but poor at system architecture, debugging emergent behaviour and owning the decision when something breaks. The real pressure sits at entry level, where the traditionally simple, well-specified tasks juniors learn on are now the most easily automated.
How can I make money using AI?
Three reliable routes, in descending order of reliability: apply it inside expertise you already have, sell implementation to businesses that lack the time or knowledge to do it themselves, or build products — though a bare "AI wrapper on a prompt" has no defensible moat.
Sources and links
IMF, Gen-AI: Artificial Intelligence and the Future of Work, SDN/2024/001, 14 January 2024 — read the paper.
IPPR, Transformed by AI, Jung & Srinivasa Desikan, 27 March 2024 — read the report.
Goldman Sachs, The Potentially Large Effects of Artificial Intelligence on Economic Growth, 26 March 2023 — read the analysis.
Department for Education, The impact of AI on UK jobs and training, 28 November 2023 — read the report.
Internal: AI automation cost for UK small business · Free AI automation audit · industry pages for recruitment, accountants and trades.
Start with the audit, not the software.
Thirty minutes on a call. We map how work moves through your business, show you where the time is going, and tell you what is worth building. You keep the roadmap either way.