The Agency Dividend Compact

Doctrine in one sentence

Build an individual economic floor from broad, observable tax bases and realised public investment returns, while keeping AI adoption open, essential rents constrained and expansion conditional on sustained evidence of labour-income erosion.

Executive summary

The supplied evidence does not establish economy-wide displacement. AI use is rising, but only 5% of AI-using UK businesses reported AI-related headcount reduction, and causal attribution remains difficult. This calls for institutional preparedness, not panic or an immediate attempt to replace the wage economy.

I propose an Agency Dividend Compact with seven planks:

1. An individual monthly Agency Credit, paid without a work requirement or household means test, then progressively recovered from high individual incomes through the tax system.
2. Cause-neutral transition insurance, protected savings and practical support for displaced workers.
3. Citizen participation in realised returns from future publicly financed technology investments.
4. A gradual shift from payroll dependence towards property, capital-income and destination-based consumption tax bases.
5. An essentials shield covering housing, energy and digital-platform rents.
6. Portability, competition and worker-transition rules that permit adoption but spread its benefits.
7. Automatic warning and scaling rules based on wages, hours, productivity, tax receipts and labour’s income share, not an unmeasurable AI percentage.

The first credit should have a net fiscal envelope of no more than 0.5% of GDP and begin only when recurring revenue has been enacted and independently scored. It would be a foothold, not a universal replacement wage. Sustained structural deterioration could raise the envelope to 1.5% of GDP. A deeper contingency plan could reach 3% of GDP, but only after fresh fiscal and distributional modelling.

This package does not rely on a robot tax, speculative investment returns, permanent borrowing, international adoption restraint or preserving unnecessary jobs. If the thesis proves substantially wrong, it still leaves better transition protection, less punitive saving rules, wider capital ownership, stronger competition and a more resilient tax base.

The policy package

1. An individual Agency Credit. Every eligible adult resident would receive the same monthly payment. It would have no work test, would not depend on a partner’s income and would not fall because the recipient had accumulated modest savings. Higher-income recipients would repay some or all of it progressively through individual taxation, avoiding a separate household means test.

   Existing disability, housing and child-related support would remain distinct. Benefit interactions would require a no-loss transition so that introducing the credit did not reduce the incomes of the poorest recipients through simultaneous benefit withdrawal. The initial amount would be calculated backwards from a net cost ceiling of 0.5% of GDP after official modelling, rather than chosen as an attractive but unfunded weekly figure.

2. Transition security independent of cause. A worker suffering involuntary job loss or a substantial, sustained reduction in hours could receive capped, declining earnings insurance for up to 18 months. Eligibility would turn on observable earnings and hours, not whether an employer admitted using AI. Pilots should compare recipients with similar non-recipients on income stability, re-employment and sustained earnings.

   The UK should replace Universal Credit’s £16,000 capital cliff with a taper and a protected transition reserve. Member states should make equivalent reforms where their systems penalise precautionary saving. Training support should pay only for assessed courses or relocation and business-start costs, with outcome publication. Retraining is useful insurance, not the presumed answer to structural loss of labour demand.

3. A public-return rule and citizen capital fund. Future public finance for commercially valuable compute, models, infrastructure or technology firms should seek an appropriate return through equity, convertibles, royalties, revenue participation or repayable finance. The instrument must fit the transaction. Procurement should obtain good services, transparency and portability; it should not be used to demand unrelated parent-company equity.

   In the UK, the National Wealth Fund’s existing £27.8 billion capacity is capital already committed to public purposes, not new money for transfers. Future qualifying digital and technology investments could nevertheless contribute realised net returns to a separately reported citizen fund. At EU and member-state level, the public component of planned AI investment could use the same principle where compatible with procurement and State aid rules.

   No dividend would be paid against estimated valuations. Only realised cash returns, averaged over several years and net of losses, costs and reserves, could support an equal dividend. The tax-funded Agency Credit would therefore survive if the investment fund performed badly.

4. An adoption-neutral tax migration. The state should not try to calculate how much of a workflow is performed by AI. It should tax observable property values, consumption, personal income, corporate profits, distributions and transactions under auditable rules.

   The UK currently charges employer National Insurance at 15% above the threshold while qualifying plant and machinery can receive full expensing. The combined incentive effect requires modelling, but reliance on payroll taxation could become increasingly fragile. Any reduction in payroll contributions should be matched, pound for pound, by scored recurring revenue from a mixture of recurrent high-value property or land taxation, narrower unjustified differences between labour and capital-income treatment, and a destination-based consumption contribution that includes imported consumption.

   Recycling consumption revenue through an equal credit can make lower-income households net beneficiaries, but that result must be demonstrated by decile modelling. A compute or AI levy is unnecessary to launch the package and should remain dormant unless its base, import treatment and avoidance rules can be made credible.

5. An essentials absorption brake. Cash agency is illusory if housing, energy or platform charges absorb it. Each government should publish rent-to-income, household-energy-burden and relevant digital-market indicators alongside the credit.

   Housing policy should combine additional supply, social or public-interest provision where governments choose to fund it, recurrent taxation of high-value immovable property and land-value capture. Cash-poor property owners could defer some liability until sale or transfer, with a recorded charge, rather than receive blanket exemption.

   Energy regulation and targeted support should focus on unavoidable household costs. Existing digital competition powers should press portability, interoperability, switching and data access. If essential-cost ratios rise persistently after a credit increase, the next increment should go first to supply, competition or targeted services rather than automatically increasing cash.

6. Adoption with transition rights. Employers should consult workers when a planned reorganisation crosses observable thresholds for headcount, hours or role changes. Consultation would cover timing, redeployment, training and sharing verified productivity gains through pay, profit sharing or shorter hours. It would not create a veto over technology or a duty to retain commercially unnecessary posts.

   Public contracts can require contract-related evaluation, portability, workforce planning and skills commitments where these are verifiable and non-discriminatory. Safety, due process and human review remain separate questions. The EU AI Act’s workplace protections can help with high-risk systems, but should not be misrepresented as income policy.

7. A standing economic-agency dashboard. Governments should monitor market income, hours, labour compensation, productivity, fiscal composition, household disposable income and concentration. AI-use figures remain informative, but would not control taxes or entitlements. This permits action when the mechanism matters economically, even if firms label their systems differently or production moves across borders.

United Kingdom: first 24 months

During the first six months, the Government should publish a distribution and fiscal stress test covering mild, medium and severe labour-income erosion. ONS business AI figures should remain in the dashboard with their developmental caveat. Broader labour, income and tax measures should determine activation.

By month 12, Parliament should receive costed options for the Agency Credit, progressive tax recovery, the Universal Credit capital taper and earnings-insurance pilots. The OBR would need to score the recurring revenues, behavioural effects and benefit interactions. One-off administrative costs must be appropriated transparently because no dedicated setup pot is identified in the supplied facts.

Between months 12 and 24, the UK should:

• Replace the £16,000 savings cliff with a taper and protected transition reserve.
• Pilot cause-neutral earnings insurance in several labour markets.
• Establish the payment and tax-recovery machinery for an individual credit.
• Begin a base credit within a net 0.5% of GDP envelope only if matching recurring revenue has been enacted.
• Apply the public-return principle to new National Wealth Fund technology transactions without retrospectively changing existing agreements.
• Add verifiable portability, evaluation and transition conditions to relevant public contracts.
• Use existing digital competition powers against lock-in and exclusion.

A new general AI regulator is not required for this settlement. Tax, welfare, investment, procurement and sectoral regulatory institutions can perform the relevant functions.

European Union and member states: first 24 months

EU institutions should establish a common economic-agency dashboard and comparable stress scenarios. This would support coordination without implying that the observed labour effects are already large or uniform.

The Commission and member states should publish model templates for individual credits, protected savings and earnings insurance. Recurring payments should remain national because member-state tax and transfer systems are the faster route. Existing EU adjustment and social funds may support eligible retraining, inclusion, evaluation and administrative preparation, but should not be presented as financing permanent income.

For the public component of the planned €200 billion AI mobilisation, including support connected with proposed gigafactories, financing agreements should disclose risk, expected public return, access conditions and concentration effects. Market-compatible equity or repayment instruments should be preferred where public capital is genuinely at risk. Procurement conditions must remain connected to the contract; any ownership instrument should sit in a separate financing agreement.

EU-level priorities should be portability, interoperability, data access, open procurement specifications and scrutiny of concentration-related rents. Member states should prepare their own revenue mixes and housing or energy absorption plans. No EU-wide income promise should be made without unanimous revenue authority and a funded budget route.

Years 3 to 5 and dormant triggers

The following thresholds are proposed design values, not observed findings. They should be back-tested before enactment and adjusted only through a published process.

An early warning would occur when any two of these conditions persist for four quarters:

• Real median working-age market income is at least 5% below its pre-set five-year trend.
• Paid hours per working-age adult are at least 5% below baseline while real output per head remains within 2% of trend.
• Labour compensation’s share of value added is at least 3 percentage points below its previous ten-year median.
• Median real hourly compensation has fallen at least 5% relative to productivity over two years.

A structural activation would require three conditions, including either the labour-share or compensation-productivity condition, for eight quarters. Subject to recurring revenue and a current fiscal score, the Agency Credit’s net envelope would then rise in steps from 0.5% towards 1.5% of GDP. The next cash step would be delayed if essential-cost indicators showed that housing, energy or platform rents were absorbing the previous increase.

A deep trigger would require both a 10% shortfall in real median market income and a 5 percentage point fall in labour’s share, sustained for eight quarters without a similar collapse in output. Government would then have 90 days to present a funded plan of up to 3% of GDP for additional credit, services or both. Three per cent is a modelling ceiling, not a claim that it could replace lost mass wages.

Between years 3 and 5, successful earnings-insurance and savings reforms should expand. Failed pilots should close. Public investment dividends should begin only if realised returns permit. Payroll-tax relief should proceed only alongside replacement revenue. Member states should activate national triggers independently, so the system does not require simultaneous international action.

An incremental triggered tier should phase down over two tax years if fewer than two conditions persist for eight quarters and an independent review finds no continuing distributional break. Payments already made would never be reclaimed.

Funding and fiscal arithmetic

The Agency Credit’s annual net cost is:

payments to eligible adults, plus administration and benefit protection, minus progressive tax recovery.

The supplied facts do not include the eligible adult population, tax-base values, recovery schedule or benefit caseloads. Exact pound, euro and weekly amounts therefore cannot responsibly be calculated here. A 0.5% of GDP net envelope costs, by definition, 0.5% of GDP after recovery. The same applies to the 1.5% and 3% tiers.

The recurring funding order should be:

1. Recurrent taxation of high-value immovable property or land, with deferral rather than exemption for qualifying cash-poor owners.
2. Removal of unjustified differences and avoidance opportunities across labour and capital income.
3. Taxation of realised corporate profits and distributions under auditable rules.
4. A broad destination-based consumption contribution as the residual source, paired with the credit and protection for essential-cost exposure.
5. Realised public investment returns, but only as an additional dividend, never as forecast funding for the core credit.

The principal net payers would be higher-income recipients, owners of high-value property, recipients of substantial capital income and consumers whose additional consumption liability exceeded their credit. Firms would face no AI-specific charge merely for adopting technology. Small organisations should receive simplified administration, not permanent loopholes.

Earnings insurance, service provision, housing supply and administration require separate appropriations. Existing EU funds can help only where their rules permit. Neither the National Wealth Fund’s £27.8 billion capacity nor Europe’s planned €200 billion mobilisation is recurring fiscal income.

Permanent borrowing is unsuitable. UK debt is already close to 95% of GDP in the supplied forecast. The ordinary EU budget cannot run a deficit, and Article 123 TFEU rules out central-bank monetary financing. Temporary national borrowing during an ordinary recession is a separate fiscal-policy decision, not the funding model for this settlement.

Even a 3% of GDP credit has not been shown here to replace mass labour income. If the severe thesis materialised, maintaining previous consumption could require much larger taxation, service provision or social ownership. The missing arithmetic includes behavioural responses, migration, avoidance, property-tax yield, consumption pass-through, inflation, benefit interactions and the sustainable division between public and private consumption. That gap must remain explicit.

Political coalition and public case

The public case is straightforward: technology should be allowed to make production cheaper, but every person should retain an independent financial foothold, and public money placed at risk should earn a public return.

A plausible coalition includes:

• Workers and unions receiving transition insurance, individual income and consultation rights.
• Innovative firms avoiding an arbitrary robot tax or job-preservation mandate.
• Small businesses benefiting from sustained household demand and, when affordable, lower payroll taxation.
• Fiscal conservatives receiving funding gates, sunsets, independent scoring and no reliance on speculative returns.
• Social democrats and civic groups receiving wider ownership, protected savings and essential-cost controls.
• Consumers benefiting from portability and greater competition.

The main losers are high-value property owners, recipients of substantial capital income, protected incumbents and some higher-consuming households. Compensation should be limited and transparent: gradual tax phase-ins, property-tax deferral for genuine liquidity problems, simple compliance for small organisations and full credit payments before progressive recovery. Existing investors should not face retrospective confiscation.

The sequence matters. Governments should announce the credit, revenue package, rent protections and public-return rule together. Paying cash first invites rent capture; taxing first without a visible dividend destroys the coalition.

Durability and anti-capture design

The tax-funded floor and investment dividend must be legally and financially separate. Ministers should not fill a revenue gap by assuming high future fund returns, and fund managers should not be pressured into politically favoured investments to increase a current dividend.

The citizen fund should have an independent board, published mandates, conflict rules, audited accounts, disclosed fees and performance against a simple diversified benchmark. Dividends should use a multi-year realised-return formula after loss reserves. Beneficial interests should not be individually saleable or usable as collateral.

Every adult should receive an annual statement showing credit received, tax recovered, fund assets, realised returns and administrative costs. Visibility creates a constituency against raids and quiet dilution. No procedural device can make a UK settlement unrepealable by a future Parliament, so political ownership and transparent costs matter more than claims of legal entrenchment.

Avoidance resistance comes from diversified bases. Immovable property cannot be offshored; destination-based consumption can include imports; individual tax recovery follows declared income; public-return terms attach to specific financing agreements. Corporate and capital-income bases remain vulnerable, so annual reporting should estimate leakage and recommend rate or base adjustments. No single volatile source should finance the floor.

Administrative power should also be divided. Statistical certification, fiscal scoring, payment administration, investment management and appeals should not sit in one body. Data collection should be limited to income, hours, tax and eligibility information needed for the programme. Adverse decisions require reasons, correction and human appeal.

Legal and institutional obstacles

The UK can alter tax and welfare rules through Parliament, but permanent transfers require OBR scoring and recurring funding. Integrating an individual credit with means-tested benefits, tax recovery and residence rules would be administratively substantial. Changes to the National Wealth Fund mandate must preserve its existing capacity and risk discipline.

The UK’s sectoral AI approach is not an obstacle because the package does not require a general AI classification. Workplace safety and due process can remain with sectoral regulators.

At EU level, direct-tax measures and new own resources normally require unanimity. The ordinary EU budget cannot borrow for a permanent entitlement, and monetary financing is unavailable. Consequently, member states must remain the principal recurring-income providers unless treaty and revenue authority change.

State aid rules affect selective public investment. Public-return instruments therefore need transparent, proportionate terms. Procurement conditions must remain related to contract performance, verifiable and non-discriminatory. Equity or royalties unrelated to the purchased service require separate agreements.

The AI Act can regulate high-risk workplace systems but cannot supply income. Overloading it with distributive tasks would confuse safety regulation with fiscal policy. Eligibility, data protection, equal treatment and appeal arrangements for national credits require legal review before launch.

Failure modes, review and exit rules

The first failure mode is a false alarm. If the thesis is wrong, a large permanent transfer and associated taxes could weaken work incentives or demand more administration than they justify. The base credit should therefore receive a full review in year five. Triggered additions phase down when the stated indicators recover. Revenue measures introduced solely for a triggered tier should expire unless reauthorised.

The second is rent absorption. Governments should test whether the lowest half of households retain real disposable-income gains after housing and energy costs. If they do not for four consecutive quarters, the next expansion must shift towards supply, competition or targeted services.

The third is poor incentive design. Reviews should publish effective marginal recovery rates, changes in paid hours and movement into sustainable work. Tax recovery should be adjusted if it creates sharp income cliffs. Individual rather than household recovery should be retained unless evidence shows a superior design.

The fourth is ineffective transition spending. Earnings-insurance and training pilots should end if, after at least three completed cohorts, independent evaluation finds no meaningful improvement in income stability or sustained earnings relative to suitable comparisons. Support should not survive merely because providers have acquired influence.

The fifth is investment capture or loss. If the citizen fund underperforms its benchmark over a full market cycle, fees, mandate and management should be reviewed. A dividend simply falls when realised returns fall. The Agency Credit must not compensate the fund.

The sixth is tax-base flight or excessive price pass-through. Before every rate increase, the fiscal authority should publish revenue, avoidance, investment and distribution estimates. If realised net revenue is less than 75% of forecast for two years, no further credit increase should occur until the base is repaired or replacement revenue is enacted.

The seventh is bad measurement. Trigger series should be frozen in advance, independently certified and protected against convenient redefinition. A statistical break suspends activation until a comparable series is produced. AI-adoption estimates can inform diagnosis but never decide entitlement.

Formal reviews should occur at month 18, year 3 and year 5. The year-five decision should separately vote on the base credit, each triggered tier, transition programmes and associated revenue. Competition, portability, protected savings and sound public-investment terms may remain even if all AI-specific contingency tiers expire.

Feasibility table

The ratings reflect present institutional routes rather than desirability. UK feasibility is strongest where Parliament can use existing tax, welfare, investment and competition machinery, but weaker where permanent revenue is politically contested. EU-level recurring income is constrained by unanimity and the balanced-budget requirement. Member states are better placed for transfers, property policy and social insurance. EU institutions are strongest on cross-border competition, investment conditions and common measurement.

What is genuinely new here

The package is not a conventional universal basic income, robot tax or sovereign wealth proposal.

Its first innovation is an agency corridor: an equal individual payment, progressively recovered through the tax system, with a modest permanent base and evidence-triggered tiers. This preserves individual independence without paying the full net amount to high earners.

Its second is the two-ledger rule. Tax revenue guarantees the floor; realised investment returns fund only an additional dividend. Public ownership can therefore grow without pretending that uncertain assets pay today’s bills.

Third is the absorption brake. Cash expansion pauses when essential rents consume the gain, redirecting the next marginal pound towards housing supply, energy relief or competition.

Fourth is tax migration without technology attribution. Revenue follows property, consumption, income, profits and distributions rather than an administratively fragile judgement about whether a worker, model or verifier produced each output.

Finally, the package contains a genuine way back. Triggered tiers can expire when labour income, hours and distribution recover. Preparedness does not become an irreversible declaration that the thesis was correct.

Bottom line

The UK and Europe should not tax the label “AI”, obstruct useful adoption or promise an unfunded replacement wage. They should build the rails for individual income, public capital returns, transition security and rent control now, then scale them using observable economic outcomes.

This settlement cannot prove that a severe collapse of labour income is already affordable. It can ensure that, if one develops, governments respond through institutions prepared in advance rather than improvised job protection, debt or surveillance.