Audit the tasks in your working week
A practical worksheet for assessing AI and software exposure in your own role, including time spent, verification, exceptions and responsibilities.
An occupation score cannot tell how you spend your week. A task audit gives you a more useful starting point: a record of the outputs you produce, the decisions you make and the consequences of getting them wrong.
Record actual work before judging automation
Choose a recent, reasonably typical week. List work in units that have an observable result: “resolve unmatched invoice,” “review a proposed design,” or “explain a policy exception.” Avoid categories such as “emails” that hide the underlying purpose. Include review, coordination and correction time, not just the time spent producing a first draft.
Estimate hours for each activity and check that the total is plausible. Add infrequent but important duties separately, such as an incident response or annual review. A weekly diary can miss the responsibility that defines the role when something goes wrong.
Use this worksheet
Download the blank worksheet (CSV) . Keep sensitive employer, client and personal information out of any example you share. The worksheet is a local file; this site does not collect your entries.
| Question | Example answer |
|---|---|
| Output and time | Update one API procedure; four hours including testing and review. |
| Possible assistance | Draft changes from an approved specification; suggest affected pages. |
| Evidence needed | The documented request succeeds in the supported test environment. |
| Exception or failure | The specification omits a permission requirement; the generated example fails. |
| Human responsibility | Resolve the discrepancy with the product owner and approve the verified procedure. |
| Next experiment | Compare total time for one assisted update with a similar manually prepared update, including corrections. |
Ask whether the whole task is covered
A tool may produce the visible artifact while leaving most of the difficult work untouched. Drafting a report is different from establishing whether its conclusion is true. Generating a schedule is different from securing the resources it assumes. Record exactly which stage the tool covers and what must happen before its output can be used.
Use three provisional categories: routine automation under known conditions; assistance with substantive review; and work led by a person because interaction, physical action or accountable judgment is central. These categories can overlap at the edges. Write the reason for your label and the evidence that would change it.
Measure useful output, including rework
For an employer-approved experiment, define the quality check first. Compare the full workflow: preparing inputs, running the tool, verifying output, correcting errors and handling exceptions. Record a failed attempt as well as a successful one. A faster first draft may not reduce total effort if review becomes harder.
Do not extrapolate a single successful example to every case. Note differences in complexity, information quality and reviewer experience. If you cannot verify the output, that is a limit of the experiment, not evidence that the tool has completed the work correctly.
Turn the audit into a skill decision
Look for work that combines a meaningful share of your time with a clear gap in your capabilities. If routine production takes many hours, learning an approved tool may help. If checking the output requires subject knowledge you lack, strengthening that knowledge may be the more useful next step. If your job includes substantial coordination, document the decisions and outcomes you already own.
Keep the occupation profile beside your worksheet and mark tasks that are absent from its sample. Revisit the record after a change in responsibilities or tools. You are building a better description of your work, not calculating personal unemployment odds.
Find your occupation · Plan a career transition · Understand the published index