Audit Your Role: Task-Level AI Exposure

Job-level automation risk figures are less useful than task-level analysis. The same job title can have very different exposure depending on which tasks dominate your week. A senior marketer running brand strategy has a different exposure profile to a coordinator producing campaign assets, even though both sit under "marketing". This module walks through a four-step method for doing that analysis on your own role, with two long worked examples that show how to apply it.

Why task-level matters more than job-level

The OECD Employment Outlook 2023 chapter on AI and the labour market found that aggregate occupation-level risk scores hide enormous within-occupation variation. The OECD specifically warned that pre-LLM exposure indexes underestimate the exposure of cognitive and white-collar tasks, and overestimate exposure for manual work that requires physical adaptation. The practical implication: the index that says your job is "low exposure" may be wrong about your specific week, and vice versa. The only honest answer comes from auditing your own tasks.

Step 1: List your actual tasks (not your job description)

Most job descriptions describe outcomes, not tasks. Start with a log of what you actually do in a typical week. Aim for 15-25 discrete task types. "Prepare monthly report" is too coarse -- break it into "pull data from system", "format into tables", "write narrative commentary", "answer follow-up questions from the FD", "present to stakeholders". Each of those has a different AI exposure score. If you started the task log in module 1, you already have most of this.

A good test: if a colleague could read your task list and predict roughly how you spend a week, the granularity is right. If it reads like a CV bullet list ("manage team", "deliver projects"), it is too coarse to audit.

Step 2: Apply four filters to each task

For each task, answer four questions. (1) Pattern: is there a large body of similar prior examples that an AI could have learned from? (2) Verification: can the output be checked for correctness quickly by a non-expert, or by you in less time than producing it? (3) Currency: does the task require information that did not exist before the model's training cutoff -- this week's prices, today's call notes, last hour's ticket queue? (4) Interpersonal: does performance depend on the specific person you are working with -- their mood, history with you, internal politics, trust?

Tasks that score yes on pattern and verification, and no on currency and interpersonal, are highest exposure. Tasks with yes on currency or interpersonal are substantially protected at current capability levels. The diagonal cases -- yes on pattern but no on verification -- are the dangerous middle: AI can produce something plausible-looking that nobody can quickly check, which is where the most expensive errors live.

Step 3: Estimate time share and value share

Two columns matter. Time share: what percentage of your working week each task type takes. Value share: how much of your perceived contribution each task represents in the eyes of your manager and skip-level. These are often misaligned, and the audit forces you to see it. High-exposure tasks that are also high time-share warrant the most attention -- they are both automatable and currently consuming significant effort. High-value tasks that are low exposure are your current structural advantage and deserve more time, not less.

Step 4: Separate exposure from risk

High exposure does not automatically mean job risk. It may mean the task will be augmented (you do it faster with AI help), delegated to a junior role, batched and pushed off-shore, or genuinely automated away. The path depends on whether your employer can redeploy the time savings productively, whether your role can shift toward higher-judgment work, and whether the savings are big enough to justify reorganising. The module on augmented roles covers this in more detail; the module on conversations with your employer covers how to influence which path your role takes.

What to do with your results

Your audit gives you three outputs. First, a list of tasks to start automating yourself -- capture the productivity gain before someone else does, and use the freed time visibly. Second, a list of high-exposure tasks to actively skill away from as a primary value driver -- not by refusing to do them but by ensuring they are not the only thing you are known for. Third, a clear picture of where your irreplaceable value currently sits, which is the input to the module on conversations with your employer.

A common mistake at this point is to read the audit and panic. The audit is not a verdict; it is a map. The same map that shows where the risk is also shows where to invest your next twelve months of skill development. Use it that way.