Sign in18 roles AI is reshaping — the ones you retrain into, and the ones you already have. For each: what the job actually involves, the skills it demands, and the curriculum that trains for it.
Free to read. No signup.
Technical roles reached by writing code and shipping real projects.
Builds products on top of large language models — retrieval pipelines, assistants and AI features inside existing software. Less model training than model application: the hard part is making a probabilistic system behave reliably in production.
What this role doesTrains, evaluates and deploys machine-learning models, then keeps them working as data shifts underneath them. Sits between data science and software engineering, and is accountable for the model in production.
What this role doesTurns raw data into decisions — designing the analysis, running it honestly, and telling the organisation what it does and does not support. The output is a decision someone can act on, not a chart.
What this role doesFinds and closes the ways a system can be broken into, before someone else finds them. Covers application, cloud and identity security, plus the governance work that proves controls exist.
What this role doesBuilds and ships complete web products — interface, server, database and deployment. The generalist role most AI products are actually built by.
What this role doesBuilds systems that interpret images and video — detection, segmentation, tracking and OCR — and makes them run fast enough to be useful on real footage.
What this role doesDesigns the architecture of systems where language models take actions — tool use, memory, planning and the protocols that let agents talk to other software safely.
What this role doesBuilds and operates autonomous agents at production scale — multi-agent orchestration, long-running workflows, and the monitoring that keeps them from quietly going wrong.
What this role doesDecides what AI product gets built and why. Owns the problem, the success measures and the trade-offs — including when the honest answer is that AI is the wrong tool.
What this role doesExisting jobs, done better with AI. No code involved.
Uses AI assistants as a copywriter, strategist and analyst — producing on-brand campaigns, emails, ads and reporting at a volume that was previously impossible, without outsourcing judgment.
What this role doesApplies AI to analysis, month-end, forecasting and board reporting — with the verification discipline finance requires, because an unchecked number is worse than no number.
What this role doesRuns every function with no team to delegate to. Uses AI as the first hire they can afford — validating ideas, writing the pitch, selling, and running operations.
What this role doesHands the preparation mountain to AI — lesson plans, differentiation, resources, feedback and reports — while pedagogy, accuracy and duty of care stay with the teacher.
What this role doesUses AI across the delivery cycle — requirements, planning, risk and stakeholder reporting — and builds no-code assistants that absorb the recurring administrative load.
What this role doesPushes the admin half of selling onto AI — research, outreach drafts, call prep, follow-ups and proposals — to spend more hours actually in front of buyers.
What this role doesApplies AI to forecasting, procurement, logistics exceptions and the institutional knowledge that usually lives in one person's head.
What this role doesUses AI as a writers' room — ideas, scripts, show notes and repurposing one idea into ten assets — while still sounding like themselves rather than like a model.
What this role doesUses AI as a personal tutor rather than a shortcut — active reading, guided problem-solving and study systems that build understanding instead of replacing it.
What this role doesFive questions, about three minutes, no account. Get an honest starting point.