Methodology
Reviewed 2026-09-01
1. Establish the factual floor
The research ledger records what the IMF, OECD, ILO, and other primary institutions actually say, with dates and direct links. A model may use a concrete number only when the ledger contains its geography, date, and context.
The baseline matters because adjacent claims are easy to blur. The OECD global minimum tax is relevant to multinational taxation, but it is not an AI contribution. IMF analysis explores AI fiscal scenarios, but it does not enact the agentic-margin proposal.
2. Label the argument
Every piece is assigned a content type such as analysis, scenario, forecast, field note, opinion, or manifesto. The opening answer states whether the mechanism is evidence, inference, or proposal.
Prospective language can be forceful. It cannot convert a campaign term into current law or a stress test into a measured outcome. Corrections change the page and its modified date rather than hiding the record in a social post.
3. Test the mechanism
A proposed contribution is evaluated against attribution, thresholds, incidence, avoidance, administrative cost, international allocation, and the destination of revenue. A clever slogan that cannot survive those questions remains a slogan.
The public calculator demonstrates relationships among inputs; it does not estimate liability. Users should obtain jurisdiction-specific professional advice for real tax decisions.
Agent contribution calculator
Both views are the same arithmetic at different granularity. Advanced works on a company's attributable revenue; the three questions work on one reader's pay and task load. Neither is a forecast: every input is a dial the reader sets.
Advanced inputs: agent-attributable annual revenue (The revenue your autonomous systems materially create or protect.); estimated wage share exposed (A scenario control for the economic activity moving off payroll.); agent contribution rate (A policy dial, shown as a percentage of attributable revenue.). Advanced model: contribution = revenue * rate; exposed = revenue * wage_share; per_million = 1000000 * rate.
The three questions: How much of your work could an agent now do end to end?; What do you earn in a year?; What share of that automated value should go back to the public?. Simple model: attributable = income * task_share; contribution = attributable * rate; monthly = contribution / 12.
Scenario model only. Not tax, legal, or investment advice.
Hypothetical scenarios
Hypothetical scenario. The company does not exist, and every figure below is an input chosen to demonstrate the model, not a measurement of any real firm.
The 10 published archetypes are: regional logistics operator, national retail chain, business process outsourcer, digital advertising agency, retail bank, health insurance administrator, software platform, legal services firm, telecom operator, utility billing operator. Each one runs the same model as the calculator: attributable revenue is total revenue multiplied by the share attributed to autonomous work, and the contribution is that figure multiplied by the rate.
