AI and jobs — sector by sector

The big worry of our time

Few things stir up as much anxiety today as the effect of artificial intelligence on jobs. It isn't an abstract fear or one trade's problem: it reaches healthcare, banking, education, insurance, tourism, leisure and, of course, technology. The question “will AI replace me?” is now being asked by professionals in every sector.

Here we try to answer it with a level head and no scaremongering: sector by sector, task by task, we estimate how much AI can automate, where employment is heading and what will have to be relearned. The recurring conclusion is nuanced: AI rarely “erases” a whole sector; it recomposes it —destroying some tasks, creating others and shifting value toward judgement, relationships and oversight.

The 2033 horizon and Kurzweil's curves

We take 2033 as our horizon because it sits on the steepest part of the AI curve. Ray Kurzweil, building on his “law of accelerating returns” —the idea that computing power per dollar grows exponentially, not linearly— places around 2029 the point where AI matches human intelligence and passes the Turing test, and around 2045 the “singularity”, when intelligence multiplies far beyond the human.

2033 lands right in between: a few years past human-level AI and inside the acceleration phase, where each year delivers more capability than the last. That is why we use that horizon across every sector — and why we expect AI to take on the bulk of routine cognitive work by then, whatever the industry.

Law of accelerating returns (Kurzweil): AI capability grows exponentially; 2033 falls between human-level AI (2029) and the singularity (2045).

How to read each sector: it's a heat map; each row is a task and three axes describe it —AI automation, headcount growth and reskilling needed—. Blue marks the “most” extreme, black the opposite, yellow the middle (the legend spells it out).

💻 Technology (IT)

What we see

• AI automation reaches almost the entire IT function: in software engineering and cybersecurity most tasks hit the top level by 2033.
• But headcount doesn't collapse. In software, cybersecurity and much of data and infrastructure the trend is to grow or hold steady: AI expands what a team can do more than it replaces it.
• The price of staying in is reskilling. The need is high in development, data and service reliability; tasks that don't evolve are the ones truly at risk.
• Cuts concentrate in lower-value coordination and service delivery, where AI does the work and headcount flattens or falls.
• The winners are roles that pair judgement with AI integration: architecture, security governance and platform engineering.

🏥 Healthcare

Where AI diagnoses, where more hands are needed and where human care is irreplaceable.

What we see

• Diagnostics —imaging, lab, pathology— is where AI goes furthest: it reads and screens at human level or better, and the specialist shifts to oversight and the hard cases.
• Care that needs hands and presence —nursing, caregiving, mental health, emergencies— grows: ageing and staff shortages outweigh automation.
• Health back-office (scheduling, coding, billing) automates heavily and cuts admin headcount.
• Research and pharma speed up with AI, but demand high reskilling in data and models.
• The most human work —therapy and long-term care— is also the safest.

🏦 Banking

From the branch to the back-office: what automates and which profiles the bank gains.

What we see

• The branch and the back-office are the epicentre of cuts: tellers, service and processing automate and headcount falls.
• Risk, credit and markets move to AI, but judgement and accountability still demand highly reskilled professionals.
• Compliance and anti-fraud grow despite AI: regulatory pressure sustains those jobs.
• Private banking holds up on the trust relationship; AI assists, it doesn't replace.
• The sector reshapes: fewer branches, more data, risk and compliance profiles.

🎓 Education

Teaching stays human; creating content, grading and admin, less and less.

What we see

• Teaching stays human: relationship, motivation and the classroom don't automate; teaching headcount holds.
• What AI absorbs is the content factory and grading: creating materials and assessing hit the top level, freeing teacher time.
• Academic administration shrinks with automation.
• Demand to TRAIN people for AI explodes: reskilling and digital skills grow fast, with high reskilling.
• Teachers who orchestrate AI (augmented tutors) win; those who only deliver content lose ground.

🛡️ Insurance

Underwriting and claims automate deeply; actuarial work and relationships hold up.

What we see

• Underwriting and claims —insurance's admin core— automate deeply: fewer handlers and near-instant decisions.
• Actuarial work levels up with AI, but stays expert, highly reskilled work.
• Distribution holds up on relationships (agents, brokers), though digital acquisition grows.
• Fraud is increasingly fought with AI, which sustains those jobs.
• Net: insurance needs fewer admin hands and more data and risk profiles.

🧳 Tourism

Intermediation falls; in-person, physical work is the safest.

What we see

• Intermediation (agencies, booking) is the most exposed: AI plans and books on its own; those jobs fall.
• In-person, physical work —kitchen, service, housekeeping, guides— is the safest: AI barely touches it.
• Front desk and service partly automate (check-in, digital concierge).
• Revenue management and marketing move to AI and gain strategic weight (high reskilling).
• Tourism stays people-intensive, but the admin value shifts to software.

🎬 Sports & leisure

Content production automates; live human talent is the safest.

What we see

• Content production (editing, graphics, marketing) and game QA is the most automatable: generative AI does the bulk.
• Routine journalism (match reports, news) automates; commentary with judgement and a voice holds up.
• Human talent —athletes, artists, live shows— is the safest: people pay for people.
• Sports analytics grows and augments the coaching staff (high reskilling).
• Net: fewer hands in the creative and admin back-room, more value in data and unique talent.

Our own estimates with a 2033 horizon, in dialogue with Ray Kurzweil's forecasts and public industry reports; indicative, not a closed forecast.