Is your job at risk from AI? A role-by-role breakdown
Job titles are a poor proxy for risk. Two people with the same title can have very different exposure depending on what they actually spend their day doing. Here's what that looks like across eight common roles.
The pattern that shows up again and again: tasks that are repetitive, text-heavy, and low on ambiguity are the most exposed. Tasks that require in-person judgment, physical presence, or navigating genuine uncertainty are the least. Most jobs are a mix of both, which is why the honest answer for almost anyone is "partially," not a clean yes or no.
Copywriters and content writers
High exposure on first-draft generation, product descriptions, and routine social copy. Lower exposure on brand strategy, voice development, and the judgment calls about what actually resonates with a specific audience. The writers who feel safest are the ones who've moved up the chain toward strategy and editing rather than staying purely in first-draft production.
Paralegals
Document review, contract summarization, and legal research have all seen real AI inroads. What hasn't moved much: client-facing judgment calls, courtroom logistics, and the accountability layer, someone still has to be liable for what gets filed, and that someone currently has to be a licensed, insured human.
Customer service representatives
Among the most exposed roles for routine, scriptable interactions, password resets, order status, simple billing questions. Less exposed: complex complaint de-escalation, situations requiring a judgment call about breaking policy, and anything where a customer specifically wants to feel heard by a person rather than resolved by a script.
Bookkeepers and accountants
Data entry, reconciliation, and routine categorization are heavily automatable and mostly already automated in modern accounting software. Advisory work, tax strategy, and audit judgment remain harder to replace, since they require synthesizing incomplete information and taking on professional liability for the conclusion.
Graphic designers
Generating variations, resizing for formats, and rough concepting have gotten dramatically faster with AI tools. Client relationship management, brand strategy, and final art direction, the judgment about which of twenty AI-generated options actually serves the brief, remain squarely human work for now.
Administrative and executive assistants
Scheduling, routine correspondence, and travel logistics are increasingly handled by AI-assisted tools. What's held up: the judgment calls about competing priorities, reading a principal's unstated preferences, and the trust relationship that takes months to build and can't be automated into existence.
Software developers
Boilerplate code, routine debugging, and first-pass implementations have shifted substantially toward AI assistance. Architecture decisions, understanding what a business actually needs versus what it asked for, and debugging genuinely novel problems remain harder to hand off.
Teachers
Grading routine assignments and generating practice materials have real AI assistance available now. Classroom management, reading a room of kids, and the relationship-based work of actually teaching, not just delivering content, remain resistant, at least with current tools.
What determines your actual exposure
Not your job title. Your task mix. Someone can hold the exact same title as a colleague and face very different risk depending on whether their week is mostly repetitive text work or mostly judgment calls, relationships, and ambiguity. That's the whole reason a task-based self-check tends to be more useful than an occupation lookup table.
Take the 2-minute task-based assessment
Frequently asked questions
My role isn't listed here. Does that mean it's safe?
Not necessarily. The task-based assessment linked above works for any role, since it looks at what you actually do day to day rather than matching your job title against a list.
Is any job completely safe from AI?
Nothing's guaranteed permanently safe given how fast the tools are improving, but roles built around physical presence, high-stakes judgment under uncertainty, and deep trust relationships have held up better than purely repetitive or text-based work so far.
← Back to the assessment · Related: The tasks AI still can't do