AI Task Exposure Auditor · Score your own work

Which of your tasks does AI absorb first?

A structured self-assessment. Score each of your tasks on five published dimensions — three that measure how exposed it is to automation, two that measure how much judgement it carries.

The map sorts your work into four honest verdicts: what AI absorbs first, what you produce with AI but sign by name, where judgement holds, and what is simply low priority. The whole model is on this page — and nothing leaves your browser.

Open the worksheet ↓ See the model first →

Published · always visible

The model, in full.

Nothing here is hidden. Every task is scored zero to four on five dimensions. Three set the automation-exposure axis; two set the judgement-weight axis. Here is the whole rubric, and the formulas that turn it into a verdict.

Axis X · Automation exposure

How much a model can take over

The higher these run, the more of the task a capable AI can simply absorb.

Standardisation — rule-based and repeatable, or novel each time
Data accessibility — inputs digital, structured, pasteable, or context-bound and tacit
Output verifiability — correctness cheaply checkable, or expensive to confirm

Axis Y · Judgement weight

How much of you it still needs

The higher these run, the more the task carries your name and your exposure.

Accountability — someone answers for this by name — professional, legal, or board exposure
Stakes — the consequence of an undetected error

The rubric — 0 to 4 anchors for every dimension
Dimension	0	1	2	3	4
StandardisationAutomation	0Novel every time	1Mostly bespoke	2Recurring shape, case-by-case	3Largely templated	4Fully rule-based & repeatable
Data accessibilityAutomation	0Tacit, in-the-room	1Scattered, needs chasing	2Mixed sources	3Mostly structured / exportable	4Digital, structured, paste-ready
Output verifiabilityAutomation	0Unknowable for months	1Expert & expensive	2Checkable with effort	3Cross-checks a known source	4Ties out exactly, instantly
AccountabilityJudgement	0No personal exposure	1Team-level, diffuse	2Named internally	3Named to board / audit / regulator	4Personal legal / professional liability
StakesJudgement	0Trivial, reversible	1Minor rework	2Material, internal	3External restatement / loss	4Regulatory / solvency / reputational

How a score becomes a verdict

Automation exposure (x) = mean(Standardisation, Data accessibility, Output verifiability)

Judgement weight (y) = mean(Accountability, Stakes)

Split Each axis runs 0–4 and divides at 2.0.

Boundary rule. A score of exactly 2.0 counts as high on both axes — ties escalate toward judgement, never toward automation.

The worksheet · your tasks

Score your work.

Click a score from zero to four for every dimension. The map and the verdicts update as you go. Add your own tasks — everything you enter is saved in this browser, and nothing is sent anywhere.

Task worksheet

Add your own task

The exposure map

Each task placed by its two axis scores. Markers sit at their true coordinates; where they collide, the number labels step aside — the points never move.

By verdict

Your tasks grouped by quadrant, each with what it implies for how you work.

The working, in prose

Method & limitations.

This is a heuristic for thinking clearly, not a machine for predicting. Read it for what it is — and for what it deliberately refuses to claim.

01A heuristic, not a prediction

This is a structured way to think, not a forecast. It will not tell you when a task will be automated — only how exposed it looks today against five explicit tests. Treat a task’s position as a prompt for a decision, not the decision itself.

02The scores are judgements

You assign every number, and reasonable people will differ. That is not a weakness to be engineered away — it is the exercise. If scoring a task forces you to argue about how verifiable its output really is, the tool has already done its job.

03Your context sets the scores

The same task sits in different places at different firms. A reconciliation with clean source systems is not the reconciliation done from twelve inboxes. Score the task as it is for you — not as a category.

04Two people, two maps — good

When colleagues score the same task differently, do not reconcile it away. The gap usually marks a real disagreement about accountability or data. Surface it, and you have found something worth governing.

05An empty corner is a finding

Few core finance tasks land in “Low priority” — low exposure and low stakes at once is rare. If several of yours do, ask why they are on your plate at all. If none do, that is telling too.

This is general information and a structured self-assessment — not advice.

Advisory · the facilitated version