PUBLIC METHOD · VERSION 1.2
Understand what we measure.
HumanWork Advantage connects your allocation of time with documented assessments of AI capability, task by task. The method addresses the work performed and its operating conditions.
The starting point
The September 2026 KMS reference contains 20 families and 223 tasks. Each family totals 100% of reference time. These proportions initialise the questionnaire; your declaration fully replaces them in the calculation. A difference of 10 percentage points or more is flagged without automatic correction.
A missing task is not added automatically. If part of your activity is absent from the list, diagnosis is paused until the reference is updated.
The capability scale
Levels describe capability in real conditions, accounting for access, cost and acceptability. Gain ranges belong to the reference definitions; they are not a promise of individual productivity. The scale is not a continuous schedule of time savings.
| Level | Name | Definition |
|---|---|---|
| N0 | Out of reach | No demonstrated capability for the core of the task under real working conditions. |
| N1 | Marginal assistance | Peripheral support: preparation, documentation or reminders. The core task remains human. The reference places the indicative gain below 20%. |
| N2 | Substantial augmentation | AI prepares an initial version that the professional revises, checks and takes responsibility for. The reference places the indicative gain between 30% and 60%. |
| N3 | Supervised near-automation | Production of an acceptable deliverable within the documented scope. Human review uses sampling rather than systematic reworking. |
| N4 | Automation | End-to-end execution within the documented scope, without supervision of each occurrence. |
Documented revisions
Revision 1.1 of 12 September 2026 explicitly separates workflow autonomy from time savings. N2 documents a substantial contribution with human review and interpretation; it does not imply a 30–60% gain. The eight decisions in this revision preserve tasks, weights and original wording in the initial register. Each published assessment shows its scope, source, date and explanation. Evidence for one tool or setting does not establish universal capability.
The three zones
- Automatable — N3 and N4. Consider what can be delegated, within the task’s documented constraints.
- Augmentable — N2. Strengthen the ability to guide, revise and interpret assisted work.
- Protected — N0 and N1. Build a human contribution that is currently central. This zone can evolve; it is not a permanent guarantee.
The weighted exposure index
Index = rounded Σ (declared time in % × numeric level ÷ 4)
The coefficients are 0, 0.25, 0.50, 0.75 and 1 for N0 through N4. Declared time must total exactly 100%, with at most one decimal place per task. The calculation uses unrounded values, then rounds the final index to the nearest integer, with half units rounded upward.
This linear weighting reproduces all twenty comparative indices in the design document after rounding. Numerical spacing is a reading convention; it does not establish constant empirical distances between levels. An index of 100 corresponds to all declared time at N4; zero corresponds to all time at N0. The index measures neither a person’s value nor their professional future.
What a threshold means
A threshold describes an observable capability to reach within a precise scope. It may be technical, physical, legal or relational. It is distinct from announcing a tool. A recorded crossing includes a date, public source, explanation and destination level.
The document contains remaining thresholds, one partially crossed threshold and thresholds already crossed. An update can change zone or, for example from N3 to N4, remain in the same zone. The interface distinguishes these situations. A documented correction may revise a threshold following an explicit decision. The original wording is retained and the correction is not a historical crossing.
Evidence and result availability
The initial reference contains expert assessments, without a reproducible protocol or accompanying public sources for its levels. It is imported as an unpublished internal foundation. To make a level visible, an administrator must document a dated public source, its scope and limitations, and an explanation in both languages. The level, its explanation and dates remain accessible together.
During early access, the map remains available for every reference family. Undocumented time is shown separately and receives no level. The three zones use their share of the entire week, without renormalisation. The overall index appears only when 100% of declared time is sourced. Action suggestions use documented assessments only. Zero-time tasks do not contribute to the calculation.
Evidence statuses
Documented: a level with explanation, dated source and stated scope. Partial documentation: useful material that cannot establish a current level. Research in progress: a tracked task with no available review. Under consolidation: insufficient, inaccessible or contradictory evidence. These statuses are not probabilities or confidence percentages.
Subscriber monitoring
The subscription follows the latest saved map. When a crossing is recorded, active subscribers who allocated time to that task receive a history entry. Changes are available in the dashboard. Automatic emails are not included in the initial launch. An email is queued if delivery and notifications are enabled. A check before sending excludes removed tasks, inactive subscriptions and disabled notifications.
The time shown in history is the allocation recorded when the crossing was logged. Evidence monitoring also retains tasks without published levels. A new assessment since the saved map is highlighted without calling it a threshold crossing. Crossings concern defined remaining thresholds only. An administrator records the findings; the MVP does not discover changes on the Internet automatically.
Explore my allocation →