Artificial Idea | AI careers · practical prompts · no hype Monday, August 18, 2025 · Issue #5 · Jobs

Everyone wants a straight answer. Here is the closest thing to one that the data currently supports.

This is the piece people have been asking for since we launched.

Not the macro picture ,the 74% versus 14% gap, the bifurcating labour market, the types of roles under pressure. That framing is useful context. But what professionals actually want to know, the question underneath every article they share and every LinkedIn post they argue about, is simpler and more personal than any of that.

Is my job safe?

It is a reasonable question. It deserves a serious answer. And the honest version of that answer is more nuanced than either "yes, you're fine" or "start retraining immediately" ,because automation risk is not a binary. It is a spectrum, and where a specific role sits on that spectrum depends on a specific set of characteristics that can be evaluated, profession by profession, with reasonable precision.

That is what this issue does.

How to read automation risk correctly

Before the profession-by-profession breakdown, one methodological point that changes how you interpret everything that follows.

Researchers who study automation risk ,most influentially, the Oxford Martin School, whose work on this has become the de facto framework ,do not assess whether a job will be automated. They assess the proportion of tasks within a job that are technically automatable with current or near-term AI capability. Those are meaningfully different questions.

A job that is 60% technically automatable is not a job that will disappear. It is a job that will change ,where the 60% gets absorbed by tools and the remaining 40% becomes the core of what the human in that role is paid to do. Understanding this distinction is critical, because it reframes the question from "will I lose my job?" to "which parts of my job will change, and am I positioned well for what remains?"

With that framing in place ,here is where the major professional categories sit.

High exposure ,significant portions technically automatable

Data and financial analysis (routine) Roles focused primarily on gathering data, formatting reports, producing standard financial summaries, and generating templated analyses are facing the highest exposure of any white-collar category. The Oxford Martin School's updated 2025 estimates put routine data processing roles at 78% technical automatability. The qualifier "routine" is doing important work in that sentence ,analyst roles that involve genuine interpretation, client communication, and strategic recommendation are in a substantially different position than roles that primarily produce standardised outputs from structured data.

Customer service and support (tier one and two) First and second-tier customer service ,handling queries that follow predictable patterns with resolvable answers ,is already being automated at scale across financial services, telecommunications, and e-commerce. The roles that remain are escalation handlers, complex case managers, and relationship managers: positions that require judgment, empathy, and the ability to navigate genuinely ambiguous situations. If your customer service role is primarily script-based, the exposure is real and present. If it involves significant relationship management or complex problem-solving, the picture is considerably less alarming.

Content production (templated) Writers, editors, and content producers whose primary output is templated, SEO-driven, or formulaic ,product descriptions, standard blog content, press release drafts ,are operating in the most disrupted content category. This does not extend to all writing. Long-form analysis, original research, narrative journalism, creative work with a strong individual voice ,these categories are significantly less exposed, both technically and from a market demand perspective.

Moderate exposure ,changing faster than most people expect

Legal (junior and paralegal) Junior legal roles focused on document review, contract comparison, basic research, and template generation are experiencing significant compression. Large law firms are already running pilot programmes that use AI to do in hours what teams of paralegals previously did in weeks. The compression at the junior end of the legal profession is real and will deepen. Senior legal work ,advocacy, complex negotiation, strategic counsel, client relationships ,is a different matter entirely.

Software development (junior) This is the category that surprises people most, given that software developers are among the most technically sophisticated workers in the economy. Junior developer roles focused on boilerplate code, standard feature implementation, and bug fixes in well-documented codebases are facing meaningful automation pressure from AI coding tools. Senior developers ,those making architectural decisions, managing complex systems, and working on novel problems ,are if anything becoming more productive and more valuable with AI assistance.

Marketing (execution-focused) Marketing roles whose primary value is execution ,producing content to a brief, scheduling campaigns, compiling performance reports, managing templated communications ,are under measurable pressure. Strategy-focused marketing roles, brand leadership, and work requiring genuine creative originality are considerably more resilient.

Lower exposure ,structurally resistant to current AI capability

Healthcare (clinical) Doctors, nurses, physiotherapists, and clinical psychologists are among the most AI-resilient professionals in the current labour market. The combination of physical presence, real-time judgment under uncertainty, legal accountability, and the irreducible human dimension of clinical care creates a set of requirements that current AI cannot meet. AI is entering healthcare rapidly ,but as a tool that augments clinical professionals, not as a replacement for them.

Skilled trades Electricians, plumbers, HVAC technicians, civil engineers working on-site, and construction managers operate in physical, variable, unpredictable environments that remain deeply resistant to automation. There is a reason applications to trade programmes are rising sharply among younger workers ,not because trades are glamorous, but because they are, by the nature of the work, among the most durable career choices available right now.

Education and coaching Teaching, coaching, mentoring, and facilitation are resilient for the same reason clinical healthcare is resilient: the human relationship is not incidental to the value ,it is the value. AI is entering education as a tool, and it will change how teachers work. It will not replace the function that a skilled teacher performs for a student who needs more than information.

Senior management and strategic leadership Roles whose primary function is judgment under uncertainty, organisational leadership, complex stakeholder management, and long-range decision-making remain highly resistant to automation. Not because these roles cannot be informed by AI ,they can and increasingly will be ,but because the accountability, the relationship capital, and the institutional trust embedded in senior leadership positions are not technically replicable.

The most important thing this breakdown tells you

No profession sits entirely in one category. Every role, at every level, contains a mix of high-exposure and low-exposure tasks. The professionals who will navigate this transition most successfully are not necessarily those in the "safest" professions ,they are those who, regardless of profession, identify which parts of their specific role are automatable and proactively own the augmentation of those parts before someone else decides it for them.

The radiologist who learns to use AI diagnostic tools is not replaced by those tools. They become the professional who can interpret AI-assisted imaging faster and more accurately than a radiologist who refused to engage. The junior lawyer who learns to use AI document review tools does not become redundant ,they become the junior lawyer who can do the work of three, which has obvious implications for their career trajectory relative to peers who are still doing it manually.

The through-line across every profession, at every exposure level, is the same. Engagement is protective. Avoidance is not.

The action

Pull up your job description ,or write down what you actually do day to day, which is usually more accurate. Go through it task by task and apply the framework from this issue. High exposure, moderate exposure, low exposure. Be honest. The tasks you flag as high exposure are not threats ,they are the first items on your AI learning list.

Thursday's issue gives you the specific prompt technique for turning that list into a personal upskilling plan in under thirty minutes.

The answer to "is my job safe" is almost never yes or no. It is: here is what is changing, here is what is not, and here is what to do about it. That answer is less satisfying than a headline. It is also considerably more useful.

,The Artificial Idea team