Everybody has an opinion on AI and jobs. Most of those opinions are either catastrophising or minimising, and almost none of them are specific. Let me be specific.
(Nobody else seems to want to. You’re welcome.)
What “Replaced” Actually Means
Before you panic or relax, let’s be precise. “Replaced” here means: a company deployed a tool and reduced headcount in a specific role as a direct consequence. Not “augmented.” Not “transformed the role.” Headcount down, tool running, cost saved.
This happened. It’s happening. Here’s where.
The 7 That Actually Did It
1. AI Customer Support Agents (Intercom, Zendesk AI, Freshdesk AI)
This is the clearest and largest-scale displacement of 2024–2025. Customer support tier-1 roles — the people answering repetitive questions about order status, account access, basic troubleshooting — are being replaced at scale.
Klarna announced in early 2024 that their AI assistant was doing the work of 700 human agents. That number is now higher. Dozens of mid-market companies followed the same playbook without the press release.
The tools: Intercom’s Fin, Zendesk AI, Freshdesk’s Freddy. These are not chatbots from 2018. They resolve the majority of tier-1 tickets without escalation, 24/7, in multiple languages.
What survived: escalation handlers, complex case managers, customer success for high-value accounts. Tier-1 volume support? The math stopped working for humans.
2. AI Copywriters for Performance Marketing (Jasper, Copy.ai, Anyword)
Junior copywriters producing A/B test variants for paid campaigns have largely been replaced at companies running volume performance marketing. Not “assisted.” Replaced.
The economics are brutal: an AI produces 50 headline variants in 30 seconds at zero marginal cost. A human produces 5 in an hour at €30–50. If you’re running 10 campaigns simultaneously, the math isn’t close. (It never was. We just didn’t say it out loud.)
Senior copywriters with brand strategy responsibilities? Still employed. Junior copywriters doing variant generation? The role effectively doesn’t exist anymore at companies that figured this out.
3. Data Entry and Document Processing (UiPath + AI, Microsoft Copilot, Rossum)
Accounts payable clerks, invoice processors, data migration roles — these are being automated by the combination of RPA (robotic process automation) and modern AI document understanding.
Rossum specifically handles invoice and document extraction at a level that made entire AP departments redundant at companies processing high volumes. This isn’t new — it’s been building since 2022 — but the scale of deployment accelerated significantly in 2025.
4. Code Review and Testing (GitHub Copilot, Cursor, Devin)
Junior QA engineers running manual test cases and entry-level developers writing boilerplate code have seen meaningful displacement. Not elimination of software engineering — significant reduction in junior roles that previously served as the entry point to the profession.
Devin, the AI software engineer, handled full end-to-end implementation tasks in documented case studies. Whether it “replaced” engineers at scale is debatable — many companies are running smaller engineering teams to build equivalent or larger products.
5. Content Moderation at Scale (Various, proprietary)
Social platforms and marketplaces that previously employed large moderation teams have dramatically reduced those teams through AI-assisted or AI-primary moderation systems. This happened quietly because it’s not good PR, but the headcount numbers don’t lie.
6. Financial Report Summarisation (Bloomberg AI, proprietary tools)
Analysts at financial institutions who previously produced standardised summaries of earnings reports, regulatory filings, and market data have been partially displaced by AI summarisation pipelines. The surviving roles are senior analysts providing interpretation and investment thesis, not the people distilling 80-page 10-Ks into two-page summaries.
7. Recruitment Screening (HireVue, Greenhouse AI, Workday AI)
Initial CV screening and first-round interview scheduling, which previously required recruiter time, is now largely automated at scale at companies using modern ATS platforms with AI screening. The entry-level recruiter role that consisted primarily of managing this pipeline has contracted significantly.
The 5 That Didn’t (Despite the Headlines)
1. Radiologists
The “AI will replace radiologists” headline has been running for eight years. It has not happened. AI is a diagnostic aid that makes radiologists faster and more accurate. The liability framework, the edge cases, and the regulatory environment mean a human physician remains in the loop. Will this change? Probably. Has it? No.
2. Lawyers (Mostly)
AI dramatically accelerated contract review and legal research. It did not replace lawyers. It made good lawyers more productive and exposed the value gap between average lawyers and excellent ones. If anything, it reduced the need for large associate teams at law firms without reducing the demand for senior counsel. A more unequal profession, not a smaller one.
3. Graphic Designers (Senior Roles)
Midjourney and DALL-E ate the stock photo industry and displaced designers producing templated, low-creativity assets. Senior designers with genuine creative direction capabilities are if anything more in demand, because the ceiling on production quality has raised and clients expect more. The junior pipeline is damaged. The senior profession is fine.
4. Teachers and Educators
AI tutoring tools are genuinely impressive. They have not replaced teachers in any meaningful scale. The roles are more social and contextual than the “AI solves problems” narrative accounts for. This will be slower than predicted. (Shocking news: turns out children are complicated. More at 11.)
5. Software Architects
Despite Devin and the various “AI developer” narratives, senior software architecture — the decisions about system design, infrastructure trade-offs, long-term technical strategy — is not being replaced. It may be the most AI-proof role in the technology industry, because the decisions are irreducibly complex and the cost of getting them wrong is catastrophic.
What This Actually Means
The pattern is clear if you look at it honestly: AI replaced roles that were defined by volume, repetition, and relatively low complexity. It did not replace roles that require judgment under uncertainty, human relationships, or accountability for high-stakes decisions.
The question to ask about any role is not “can AI do this?” — AI can produce something that looks like almost anything. The question is “what does getting this wrong cost, and who is accountable?” Where the answer is “a lot” and “a named human,” AI is an accelerant, not a replacement.
Where the answer is “not much” and “it doesn’t really matter who,” you should probably update your CV.
Stay sharp.
The best AI analysis, straight to your inbox.