Agentic economy
From useful intelligence to shared prosperity.
An agent can do work. The harder questions are who authorises it, who remains responsible, and who receives the gains.
What makes an economy agentic?
An AI assistant may prepare a recommendation. An agent connected to tools can also carry out a sequence of actions within delegated authority: reconcile records, prepare a filing, compare quotations or initiate an approved transaction. That changes how work is organised. It does not, by itself, give software legal personality, ownership or a right to contract.
The research programme separates five questions: ownership, operational control, authority to act, liability, and entitlement to returns. A system can possess a technical credential without possessing legal authority. A company can become more productive without distributing more income to the people around it.
The Manx proposition
The Island has experience organising legal identity, registration, professional services and international customer relationships. The proposed next step is to test whether that experience can make delegated AI activity easier to verify and govern.
Begin with bounded workflows: company-service administration, maritime and aviation records, and permissioned data products. Measure the full cost of integration, verification, supervision, correction and failure. Compare against simpler digital tools and established human processes.
Production and participation are different tests
More accepted output is one outcome. Better wages, capital receipts, useful public services and retained human authority are others. AI access can increase capability; cash transfers provide purchasing power; dividends require funded receipts and enforceable entitlements. These instruments should be compared rather than bundled into a single promise of abundance.
The academic work develops a conditional argument: productive renewal can increase output while weakening the employment channel through which households participate in its gains. Demand, new tasks, ownership and institutions influence the result. It is a research question, not a forecast that unemployment is inevitable.
Three experiments worth testing
- Administrative preparation: retrieve published requirements, identify missing documents and prepare drafts for an authorised person.
- Asset-service operations: reconcile maintenance or mortgage records and flag discrepancies. Safety certification stays with the competent people and authorities.
- Governed data products: document provenance, permitted AI uses, quality, update obligations and the lawful licensor before seeking customers.
Read the developed strategy → · Explore the academic argument →