How to Actually Evaluate Whether AI Will Affect Your Job

The standard approach to AI job risk analysis works like this: researchers take a list of occupations, score each one on how many of its tasks AI can theoretically perform, and publish a ranked list of "at risk" and "safe" jobs. Telemarketers and data entry clerks rank near the top. Surgeons and clergy rank near the bottom. You look up your job title, find a risk percentage, and either feel reassured or anxious.

The problem is that this approach answers the wrong question. It tells you whether AI can theoretically automate tasks associated with a job title. It does not tell you whether AI will actually displace you, or when, or what specifically makes you more or less exposed within your occupation.

Yale's Budget Lab director Martha Gimbel put it cleanly: "We do not have a good track record of predicting how technological change will play out in the labor market."[1] ATMs were supposed to eliminate bank tellers. Employment of bank tellers increased after ATMs were introduced, because branches became cheaper to operate, so banks opened more of them. Radiologists were supposed to be replaced by AI diagnostic tools. Radiology employment has grown. The predictions were not wrong about what AI could do. They were wrong about how organizations, labor markets, and human behavior adapt around new capabilities.

There is also a critical distinction in the research that most popular coverage glosses over, and it is the most important thing to understand before applying any risk framework to your own situation.

The most useful reframe in all of the AI labor market research comes from the Goldman Sachs analysis: the question is not "can AI do my job?" It is "what fraction of my actual daily work involves tasks where the correct output can be determined algorithmically from structured inputs?"

That question sounds technical but it is not. You already know the answer for your own work. Think about the last five things you did today. For each one, ask: could someone write a complete rule for how to do this correctly, in advance, without seeing the specific instance? If the answer is yes, that task is automatable in principle. If the answer is no, because the right response depends on context, judgment, relationships, or information that cannot be fully specified in advance, it is not.

A data entry clerk transcribing handwritten forms into a database is doing a task where the correct output can be completely determined by a rule. AI does this better than humans already. A therapist deciding how to respond to a client who just disclosed something unexpected is making a judgment call that cannot be reduced to a rule. No current AI system comes close to replicating it.

Most real jobs contain tasks of both types. The question for your specific situation is the ratio, and where the tasks that make you valuable to your employer fall on that spectrum.

Goldman Sachs Research, in its analysis of over 800 occupations, identified four factors that best predict whether AI will displace specific work.[2] Brookings Institution's research on adaptive capacity adds a fifth that the Goldman Sachs analysis underweights.[3] Here are all five, with honest explanations of how to apply each one.

Work through these five questions honestly. You will get a personalized risk profile at the end, with specific guidance based on your answers. This is not a score to panic about. It is an accurate picture of where you are exposed and where you are not, so you can make deliberate decisions about what to build toward.

The same framework applied to different roles produces very different conclusions, even within the same industry. This table works through several examples to show how the factors interact.

The framework above addresses direct risk: will AI automate the tasks in your role? But there are two second-order risks that are equally real and get almost no attention in the popular discourse.

The demand compression risk

Your job may not be directly automatable, but the industry or organization that employs you may shrink because AI automated the work of your colleagues. A senior strategy consultant at a firm that uses AI to do the research and analysis work that used to require 10 junior analysts is still employed, but the firm is smaller, growing more slowly, and hiring less than it used to. The career ladder is shorter. The opportunities for advancement are fewer. The compensation growth that came from expanding the practice is limited.

This is not the same as losing your job. But it is a real consequence of AI disruption that the "AI won't replace knowledge workers" reassurance does not capture. The structural environment around you changes even when your specific role does not.

The entry-level collapse risk

This is the risk that nobody is talking about honestly. The traditional career path in most knowledge work professions runs through a junior tier that involves a lot of the work AI is automating fastest: document review, research aggregation, first drafts, data analysis, report preparation. When those entry-level roles shrink because AI handles the work, the pipeline for the next generation of senior practitioners compresses.

As Thomas Davenport, professor at Babson College and author of The AI Advantage, put it: "If companies don't hire entry-level workers today, how do you get experienced workers tomorrow? We still haven't figured that out."[4] If you are mid-career, this may not affect you directly. If you are early-career or entering a field now, it is the most important risk to understand. The training ground is changing faster than anyone has a good answer for.

An honest assessment of your AI exposure, like the one from the interactive quiz above, is only useful if it informs action. Here is what the research suggests is actually worth doing, separated from the generic advice that gets recycled endlessly.

If your assessment says high risk on multiple factors

The useful response is not to wait and hope the disruption is slower than expected. It is to deliberately shift the ratio of your work toward the tasks that score lower risk on the framework: more judgment, more context dependency, more consequence-bearing work. In practice this usually means moving toward client-facing roles, strategic functions, or roles that require managing the AI systems rather than executing the tasks they automate. The window to make that shift proactively is shorter than it feels.

If your assessment says medium risk

The honest response is to identify which specific tasks in your role score high risk and build AI fluency for those tasks specifically, not general "AI literacy" in the abstract. If you understand the tools that are automating your most exposed tasks, you become the person who manages and improves those tools rather than the person those tools replace. PwC's research found that workers with demonstrated AI skills earn 56% more than peers in the same roles without them.[6] The gap is real and the window to develop it is now.

If your assessment says low risk

Low direct automation risk does not mean no AI relevance. The tools available to people in low-risk roles are still improving your leverage. A therapist who uses AI to reduce their administrative burden does more therapy per hour. A plumber who uses AI to optimize their scheduling and quoting spends more time on billable work. The frame shifts from "will AI take my job?" to "how do I use AI to do more of the high-value parts of my job?"

On adaptive capacity specifically

Whatever your task-level risk assessment shows, building adaptive capacity is worth the investment independently. Financial cushion, transferable skills, and a genuine professional network are valuable whether or not AI disrupts your specific role, because they reduce the cost of any career transition, voluntary or not. The Brookings research framing is useful here: exposure and adaptive capacity are two different dimensions. High exposure with high adaptive capacity is a manageable situation. High exposure with low adaptive capacity is the genuinely difficult one.