The Civic Agency Compact

Doctrine in one sentence

Let productivity rise, tax broad and observable bases rather than the presence of AI, give every person a cash and capital claim on prosperity, and prevent essential rents from swallowing the gain.

Executive summary

The response should not be a robot tax, a promise to save every job, or an immediate leap to an unfunded universal basic income. It should build a successor circuit in stages.

The Civic Agency Compact has seven planks: a universal but initially modest Agency Dividend; temporary earnings insurance for abrupt transitions; a gradual shift in social contributions away from payroll alone; public ownership claims when public money finances compute and other productive assets; an essential-cost firewall covering housing, energy and digital gatekeepers; competition, portability and public-compute access for smaller firms; and a statistical trigger system based on labour income, hours, wages and consumption rather than disputed estimates of AI use.

The United Kingdom should establish the payment and monitoring infrastructure first. An illustrative starting dividend of £600 per adult and £300 per child each year would cost £6 billion for every ten million eligible adults and £3 billion for every ten million eligible children, before tax recovery and administration. The actual eligible populations, distributional effects and behavioural responses must be modelled by the Treasury and scored by the OBR. The payment should be individual, monthly, disregarded for Universal Credit and its capital test, and progressively recovered from higher incomes through the tax system.

EU institutions should concentrate on common measurement, competition, AI rights, procurement standards and return-bearing public investment. Recurring income must initially remain a member-state responsibility. An EU-wide dividend or tax would face unanimity, national ratification for a new own resource, a balanced EU budget and treaty constraints. Those obstacles should be stated, not wished away.

If the structural risk does not materialise, most of the package remains useful: better automatic stabilisers, less tax bias against employment, portable rights, stronger competition, public investment discipline, affordable essentials and evidence suitable for future decisions. The expensive elements remain dormant until published indicators and identified funding justify them.

The policy package

1. A universal Agency Dividend. Every eligible adult should receive the same monthly payment, with a lower child payment made to the responsible guardian. It should not depend on unemployment, disability, job search, occupation or proof of AI displacement. It should be paid to individuals rather than households, remain available to workers and savers, and be disregarded for means-tested benefit withdrawal and capital limits. That avoids reproducing Universal Credit's 55 per cent earnings taper or its effective penalty on savings above the current capital limit.

The dividend should be taxable or paired with an explicit income-tax recovery mechanism, preserving universality at the point of payment while reducing the net gain at high incomes. The first rate should be modest. Its purpose is to establish a trusted pipe, a legal entitlement and a small independent income, not to pretend that £50 a month replaces a wage. Later increases should occur in fixed, pre-costed steps.

2. Transition earnings insurance. A worker suffering a large involuntary fall in earned income should receive a declining top-up based on verified payroll or tax records for up to two years. One illustrative design would replace half of an eligible loss in year one and one quarter in year two, subject to a cash cap. Self-employed people would use filed tax returns, with anti-manipulation rules. The rate, cap, qualifying fall and treatment of volatile earnings require modelling.

This insurance follows the person, not the old job. Firms receive no subsidy for preserving unnecessary posts. Recipients may take a lower-paid job, reduce hours, retrain, start a business or provide care. It therefore assists adjustment without making non-adoption commercially rational.

3. A social-contribution switch. Payroll should cease to bear an ever larger share of social financing if paid human hours contract. The switch should be gradual and revenue-neutral at each Budget. The UK should first reduce anomalies between taxation of employment, self-employment, dividends and realised capital gains, while protecting genuinely small incomes. It should then examine a broader contribution across personal income, distributed profits and property or land rents.

A dormant consumption-based supplement may be needed if labour income contracts substantially. A VAT-based mechanism has an observable domestic consumption base and established import treatment, unlike a levy on an alleged AI share. It is regressive in isolation, so every pound raised by an Agency VAT supplement should be recycled through the per-person dividend, with distributional modelling published before activation. EU member states would have to operate within EU VAT law. There should be no claim that an EU-wide supplement can be adopted without the required unanimity.

4. Citizen returns from public investment. When government supports commercially valuable compute, chips, infrastructure or firms, the default instrument should be an equity stake, preferred share, convertible loan, royalty or repayable advance where valuation, risk and law permit. Grants may still be justified for pure public goods, research spillovers or security, but the reason should be published.

The UK's National Wealth Fund already has authority to use equity and convertibles and has £27.8 billion of capacity, including a digital and technology remit. That capacity is not free money and should not be booked as dividend funding. A separately accounted Citizen Returns portfolio should receive the proceeds of qualifying new investments. After maintaining capital and covering losses, realised returns could support the Agency Dividend.

The EU's planned €200 billion mobilisation for AI, including €20 billion for gigafactories, should likewise not be described as a cash pot available for distribution. Where public authorities bear commercial risk, support agreements should capture a proportionate public return, consistent with State aid, procurement and valuation requirements.

5. An essential-cost firewall. Cash agency is illusory if scarce housing, energy suppliers or digital gatekeepers capture it. Every proposed dividend increase should therefore be accompanied by a published assessment of housing supply, rent burdens, energy costs and platform concentration.

National and local governments should combine faster housing supply, social and affordable housing investment, protection of vulnerable energy consumers and reform of taxes on land or property rents according to domestic institutions. Digital-market authorities should use interoperability, portability and data-access powers to reduce gatekeeper rents and switching costs. If essential-cost increases absorb a material share of a dividend rise for lower-income households, the next rise should pause while supply and competition measures are strengthened.

6. Transition rights, data and open access. Large employers should report aggregate headcount, paid hours, entry-level recruitment, pay distribution and contractor expenditure by occupational group. They should not be asked to declare a fictional percentage of work done by AI. Workers should receive notice, meaningful information and a route to contest high-impact automated employment decisions. EU action should build on the AI Act's workplace duties, while recognising that its original high-risk deadlines had not yet been finally changed on the verified date.

Public contracts using AI should include contract-related requirements for evaluation, data portability, security, auditability, skills transfer and an orderly supplier exit. Requirements must remain verifiable, non-discriminatory and connected to the contract. They should not become disguised domestic preference, permanent job quotas or unrelated demands for parent-company equity.

7. Competition and productive access. The CMA and European Commission should use the UK digital-markets regime and the Digital Markets Act to address cloud dependence, self-preferencing, data barriers and weak interoperability. Public compute should publish transparent access, pricing and allocation rules, including routes suitable for researchers and smaller firms. This widens the ownership and productive use of AI rather than restricting adoption to incumbents.

United Kingdom: first 24 months

Parliament should establish the Agency Dividend framework, its residency and eligibility rules, tax treatment, benefit disregard, audit requirements and staged payment formula. The first Budget should fund only the initial rate. An illustrative £600 annual adult payment and £300 child payment provides a clear costing formula without pretending that the population arithmetic has already been completed. Delivery should be tested across HMRC and DWP systems before national payment.

The Treasury should publish a social-contribution review comparing employer National Insurance, currently 15 per cent above the relevant threshold, with the treatment of self-employment, dividends, gains, distributed profits and property rents. It should recommend a first revenue-neutral switch, not simply add another tax on top of employer NIC. Full expensing should also be assessed alongside payroll taxation for neutrality between employing people and purchasing qualifying capital, without assuming that investment relief is inherently undesirable.

The National Wealth Fund should create transparent accounts for new digital and compute investments that generate public ownership or repayment rights. Ministers should publish valuation, risk, expected-return and exit rationales. No forecast dividend should count as current revenue.

ONS, DSIT, the Treasury and the OBR should agree a public labour-income dashboard. Large-employer reporting should begin with a proportionate threshold and a sunset review. Procurement guidance should add model contract clauses for portability, evaluation and contract-specific workforce transition.

Finally, government should introduce two-year earnings insurance on a controlled national basis, with rates and caps set only after fraud, distribution and cost modelling. It should reform Universal Credit so the Agency Dividend and a defined protected savings account do not extinguish support.

European Union and member states: first 24 months

The Commission should publish a common structural-transition scoreboard using harmonised labour-share, hours, wage, hiring and household-income measures. It should distinguish observation from causal claims about AI. Member states should submit contingency plans showing how their existing tax and transfer systems could pay an individual dividend if agreed thresholds were crossed.

EU institutions should enforce the AI Act duties already in force and prepare employers for the applicable high-risk rules without treating the proposed later dates as finally enacted. The Commission should also use the Digital Markets Act review to pursue AI-related interoperability, cloud portability and data access.

EU procurement guidance should clarify lawful contract-linked transition conditions, evaluation requirements and exit portability. ESF+ and the European Globalisation Adjustment Fund should support training, placement, advice and local adjustment within their mandates. They should not be presented as sources of permanent household income.

Member states should build their own dividend payment capability, earnings insurance and essential-cost firewall. They should review payroll contributions and broad domestic tax bases within EU law. Selective compute or industrial support should undergo State aid review, with return-bearing instruments preferred where commercially and legally suitable.

The Commission may prepare options for a future common own resource, but it should state plainly that direct-tax action and new own resources ordinarily need unanimity, and that the EU budget cannot borrow routinely to pay a permanent dividend.

Years 3 to 5 and dormant triggers

The compact should use a two-key trigger. The first key is structural evidence. A formal review begins if at least three of the following proposed indicators breach their thresholds for four consecutive quarters, after cyclical adjustment:

1. Labour compensation falls at least 2 percentage points below its trailing five-year share of national income.
2. The share of working-age adults mainly supported by wages falls at least 3 percentage points.
3. Real median earned income falls 5 per cent behind the productivity path over three years.
4. Paid human hours per working-age adult fall 5 per cent while real output per adult is broadly stable or rising.
5. The share of household consumption financed by wages falls at least 5 percentage points.

These are proposed decision rules, not observed conditions. Statistical agencies must back-test definitions, revisions and false positives before activation.

The second key is a fiscal and supply certificate. The independent fiscal institution must confirm recurring finance; government must publish inflation and essential-cost effects; and delivery systems must be ready. Crossing the statistical threshold authorises a pre-costed dividend step, but does not authorise borrowing without limit.

If the first step proves stable, later steps can be financed by further contribution reform and, if necessary, a one or two percentage point Agency VAT supplement recycled wholly through the dividend. Exact rates remain dormant until revenue and distribution have been modelled. Public-investment returns may supplement, but never underwrite, the commitment.

Funding and fiscal arithmetic

The initial UK dividend's gross annual cost is:

£600 × eligible adults + £300 × eligible children

Every ten million adult recipients add £6 billion; every ten million child recipients add £3 billion. Administration adds further cost. Tax recovery from higher-income recipients reduces the net fiscal cost, but the amount cannot be stated without microsimulation. The OBR must score participation, consumption, tax and benefit interactions.

An additional £100 per adult per month would cost £12 billion for each ten million adult recipients each year. At EU member-state level, €100 per month costs €1.2 billion annually for every million recipients. These formulas expose the central fact: a payment capable of replacing a large share of mass wages would require fiscal resources far beyond a narrow levy on a few AI companies.

The immediate payers should be higher-income recipients through tax recovery; beneficiaries of gaps between labour and capital-income taxation; owners of under-taxed land or property rents where domestic reform permits; and recipients of distributed profits and realised gains. Each source needs behavioural and avoidance modelling. Low-income workers, small savers and ordinary pension provision should be protected through thresholds rather than occupational exemptions.

Transition earnings insurance is a separate temporary liability. Its cost depends on the qualifying earnings fall, replacement rate, cap and incidence. No credible total is available in the supplied facts. It must be capped and annually appropriated.

Public-investment returns are uncertain, delayed and risky. Neither the National Wealth Fund's capacity nor the EU's mobilisation target is spendable revenue. A consumption supplement is the more scalable dormant base because domestic consumption remains observable even when production methods change, but its yield, inflation effect and distribution must be modelled before enactment.

The package therefore funds a modest first floor plausibly, but does not claim to have funded a future full living income. If labour income were to collapse at the strongest level contemplated by the thesis, tax rates, public spending priorities and the size of the state would require an explicit democratic settlement.

Political coalition and public case

The public argument is simple: technology should make the country more productive without making ordinary people economically powerless. Government will not tell firms to use yesterday's tools. It will ensure that everyone receives a basic claim on progress, can survive a transition, and is not forced to surrender the gain to rent, energy or platform charges.

Workers and trade unions gain an unconditional floor, earnings insurance, aggregate workplace evidence and contest rights. Innovators gain adoption-compatible rules rather than a gameable AI tax. Small firms gain interoperability, public-compute access and lower lock-in. Families gain individual payments and protection from benefit withdrawal. Fiscal conservatives gain funding certificates, staged activation and exit rules. Regions hosting infrastructure gain investment, provided support is not a disguised permanent subsidy.

The visible losers are some high-income households, owners of concentrated land and capital rents, digital gatekeepers benefiting from lock-in, and firms exploiting tax distinctions between labour and capital. Changes should be phased. Asset-rich but cash-poor property owners could be offered deferral of an added property charge until sale or transfer, with interest, rather than exemption. Firms should receive notice and stable tax schedules, not compensation for every lost preference.

Durability and anti-capture design

The Agency Dividend should have a transparent statutory formula, separate accounts and an annual statement showing gross cost, tax recovery and funding sources. Ministers should not be able to increase it without an independent fiscal certificate, or quietly reduce it through benefit withdrawal.

Citizen Returns assets should be professionally managed under a published mandate, with independent appointments, conflict disclosures, external audit, exposure limits and no ministerial direction of individual investments. Returns should be measured after losses and capital maintenance. Public support agreements and exits should be published, subject to genuine commercial confidentiality.

Broad bases reduce avoidance. Land cannot relocate, domestic consumption captures imported as well as local goods through existing border treatment, and personal tax bases can use established reporting. Connected-party rules, beneficial-ownership checks and consolidated reporting should address fragmentation. The package still needs international cooperation on profit shifting, but it does not depend on every country taxing AI.

Hostile future governments cannot be prevented absolutely, especially under UK parliamentary sovereignty. Durability instead comes from visible individual payments, distributed beneficiaries, independent accounts, automatic publication and a coalition that includes workers, families, small firms and investors in competent public assets.

Legal and institutional obstacles

In the UK, Parliament can change tax and welfare rules, but permanent spending requires legislation, Treasury funding and OBR scoring. Debt near 95 per cent of GDP makes routine debt finance an unsuitable answer. Housing and property measures interact with devolved and local responsibilities. Procurement conditions must comply with transparency, fairness and trade obligations. The UK has no general AI Act, so workplace measures must use sectoral, employment, data and procurement law or new, clearly proposed legislation.

At EU level, recurring income is principally a member-state task. Direct-tax decisions normally need unanimity. A new own resource needs unanimity and national ratification. The ordinary EU budget cannot run a deficit, and Article 123 TFEU rules out central-bank monetary financing of public authorities. Article 153 excludes EU action on pay and makes parts of social-security coordination particularly sensitive.

State aid review is required for selective national industrial support. Procurement conditions must relate to the contract and remain verifiable and non-discriminatory. Free-movement, equality and social-security rules will shape member-state residence eligibility. None of these obstacles prevents national dividends or return-bearing public investment, but they rule out presenting a rapid, centrally funded EU income as administratively routine.

Failure modes, review and exit rules

Fiscal insufficiency. The greatest risk is promising an income that the proposed taxes cannot support. Every rate increase should require ten-year fiscal projections, sensitivity tests and a named recurring source. If funding fails, the increase does not occur.

Rent capture. If, within eighteen months of a dividend rise, housing and household-energy costs for the lower two income quintiles rise by more than 30 per cent of the cash gain relative to their prior path, the next rise pauses. Government must report whether supply, regulation, market power or a common shock is responsible.

Measurement error. The structural trigger uses several indicators because none is decisive. Data revisions should not claw back payments already made. Thresholds should be reviewed every three years, with changes prospective and independently explained.

Ordinary recession mistaken for discontinuity. Cyclical adjustment and the requirement that output per adult be stable or rising help distinguish productive displacement from a general slump. Earnings insurance and existing benefits should handle recessions even when the structural trigger is not met.

Inflation or labour-supply effects. Each dividend step should receive an ex post review after eighteen and thirty-six months. If participation falls materially beyond the forecast range, or demand persistently exceeds productive capacity, the next step pauses. Recovery from high incomes, tax rates or the rate of future increases may change, but accrued payments are not reclaimed.

Fund capture or poor investment. Citizen Returns should be benchmarked against its risk and policy mandate. Persistent underperformance over five years triggers an independent review, replacement of managers and possible closure to new investments. Existing assets should be sold only through an orderly value-for-money process.

Transition programmes that do not work. Earnings insurance ends after two years per claimant. Training suppliers should be paid partly against completion and independently measured employment or earnings outcomes. Programmes with no improvement against a credible comparison after three cohorts should end or be redesigned.

Thesis substantially wrong. If fewer than two structural indicators remain breached for eight consecutive quarters, no triggered dividend increase occurs and any temporary VAT supplement should expire at its legislated review date unless renewed after a fresh distributional case. The modest baseline dividend may continue if Parliament judges its ordinary anti-poverty, savings and stabilisation benefits worth the cost. Competition, portability, public-return discipline and transition insurance remain valuable regardless.

Feasibility table

The ratings concern the ability to start a lawful, administrable plank, not certainty of political agreement or full-scale funding. UK feasibility is lower where permanent fiscal commitments require difficult scoring. EU-level feasibility is low for recurring tax and income because of unanimity, budget and treaty constraints. Member-state feasibility is generally higher for tax, welfare, housing and energy, but fiscal space varies.

What is genuinely new here

The central innovation is not any single tax or benefit. It is the combination of an agency escalator and a rent firewall.

The escalator is activated by observable deterioration in labour-financed agency, not by counting robots, tokens or AI-generated sentences. It separates statistical evidence from fiscal permission through a two-key trigger. This makes preparation automatic without making huge expenditure automatic.

The rent firewall treats distribution as incomplete until the recipient can retain the gain. A dividend increase is therefore linked to evidence on housing, energy and digital-market capture, with a pause rule if essential costs absorb too much.

The third novel element is symmetry in industrial policy. If taxpayers finance high-risk productive infrastructure, they should sometimes receive a financial claim, not merely hoped-for jobs or cheaper services. That claim is governed as capital and cannot be counted as revenue before returns exist.

Together these elements create a successor circuit: productive adoption raises output; broad taxes and public returns capture part of the gain; an individual dividend sustains agency and demand; competition and essential-cost policy stop the transfer being captured; and evidence determines whether the circuit needs to expand.

Bottom line

The UK and EU should prepare for structural labour-income weakness without declaring it inevitable. The practical route is national, staged and adoption-compatible: build a small universal payment pipe, insure abrupt earnings losses, broaden social financing beyond payroll, retain public claims on publicly financed assets, and attack essential rents and gatekeeper power.

The first payment can be costed honestly. A future wage-replacing income cannot yet be. That gap is not a reason for inaction, but it is a reason to reject slogans. The Compact creates institutions that can scale if the evidence arrives and remain useful if it does not.

Participant: Codex (GPT-5)