The Agency Circuit Compact

Doctrine in one sentence

Create an individual, portable claim on cash, affordable essentials and public upside, funded only from observable and genuinely recurring bases, while allowing AI adoption and responding to measured household outcomes rather than an unprovable AI share of work.

Executive summary

Unit Cost Dominance is a contingency diagnosis, not evidence that the wage-demand circuit has already failed. The reported figures support caution rather than certainty: in June 2026, 29% of UK businesses reported AI use, rising to 49% among firms with at least 250 employees, while 5% of AI-using businesses reported AI-related headcount reductions. ONS says the data are in development. DSIT also cautions that causal attribution is difficult. EU enterprise use was 20% in 2025. None of this justifies waiting for proof before designing institutions.

The response should not be an AI tax, a robot tax, or a programme to preserve work which is no longer economically necessary. All depend on contested definitions, invite relabelling and offshoring, and risk taxing the adoption that may raise productive capacity. Nor should public investment capacity be described as if it already finances a citizen dividend.

I propose an Agency Circuit Compact with two clocks. The first clock starts now and has ordinary public-policy value: remove damaging means-test barriers to saving and intermittent work, contain housing, energy and platform rents, improve portability and competition, make public AI procurement contestable, and build measurement and delivery capacity. The second clock is dormant: it scales an individual cash entitlement only when persistent household outcomes deteriorate and independently certified recurring funding is available.

The compact has six connected parts:

1. An individual Agency Account and, when funded, a Universal Agency Floor that does not require a claimant to prove an AI-caused job loss.

2. A rent and essential-cost firewall so extra purchasing power is not simply transferred to landlords, energy suppliers or locked-in platforms.

3. An Agency Fund that separates capital, cost-reduction investment and recurrent household payments.

4. A gradual shift away from relying so heavily on payroll taxation, but only after published modelling of alternative observable bases, incidence and avoidance.

5. Competition, procurement and due-process rules that improve state capacity and market exit without pretending that human review restores mass labour demand.

6. A two-key trigger system based on outcomes and fiscal coverage, not technology adoption.

The UK should lead on household rules, tax-and-transfer choices and National Wealth Fund governance. EU institutions should lead on cross-border competition, portability, procurement guidance and common measurement. Member states should lead on recurring income support, housing, energy and national investment vehicles. This is more credible than promising an EU-wide income entitlement despite unanimity requirements for direct tax and own resources, the no-deficit EU budget rule and the limits on monetary financing.

The policy package

The central unit is an Agency Account for each eligible adult resident, administered through existing tax and welfare channels as far as practicable. Initially it is a protected savings and capability account: its balance should not automatically disqualify someone from support, and it should allow people with variable earnings to retain a liquid buffer. In the UK, that directly addresses the present Universal Credit capital rule, which normally removes eligibility above £16,000. The precise disregard, taper interaction and eligibility rules must be costed before enactment.

The account becomes a Universal Agency Floor only when paid for. That floor should be an individual monthly entitlement, outside household means tests and not conditional on being unemployed, retraining, accepting a particular job or establishing the cause of an income loss. It is not a substitute for disability support, childcare, housing policy or public services. Its purpose is to preserve a minimum capacity to refuse abuse, move, learn, save and buy necessities when wage income is less dependable.

The package must put rent containment ahead of dividend rhetoric. Cash transferred into a constrained housing or energy market can be captured. The UK and member states should therefore pair any expansion of cash support with housing supply, local infrastructure and transparent sharing of planning or land-value uplift where domestic law permits. One-off or cyclical uplift should finance supply, infrastructure or reserves, not a permanent payment. A blanket rent cap should not be assumed to solve the problem, because it may have supply consequences. Instead, every jurisdiction should publish a rent pass-through assessment before increasing the Agency Floor.

Publicly supported energy and efficiency projects should include measurable household-cost outcomes where contractually and legally appropriate. Claimed bill savings should be measured before they are counted as a distributional gain. For digital markets, portability, interoperability, data access and viable exit from public-sector systems are anti-rent measures. They can reduce lock-in and improve bargaining power, but they are not income distribution and should never be booked as revenue for a dividend.

A ring-fenced Agency Fund should have three separate ledgers:

• A capital ledger for public equity, convertible returns and one-off receipts. It reinvests, absorbs losses and builds reserves.

• A cost-reduction ledger for housing, energy and digital-access investments with published household outcomes.

• A recurring agency ledger for stable, scored recurrent revenue that can finance the Universal Agency Floor, earnings disregards or temporary supplements.

The National Wealth Fund is relevant because it can use equity and convertibles and has £27.8bn of capacity, including a digital and technology remit. It is not, however, a pre-funded citizen dividend. Capacity is neither annual income nor distributable cash. Only realised returns, after investment obligations and prudent reserves, may enter the recurring agency ledger. The same discipline applies to EU and national AI investments: public risk may justify public returns on an investment basis, but procurement cannot be used to demand unrelated parent-company equity.

The tax side should be a contribution rebalancer, not an AI levy. The UK currently has employer National Insurance at 15% above the 2026/27 threshold, alongside permanent full expensing for qualifying plant and machinery. The right question is whether some future payroll reliance can be replaced by more robust, observable bases, not whether a firm used AI in a workflow.

Candidate bases for modelling include domestic land or planning-value capture, legally defined regulated or scarcity rents, taxable domestic business surplus and realised public investment returns. Each needs a legal taxpayer, domestic nexus, treatment of imports and foreign suppliers, group-accounting rules, anti-avoidance measures and an incidence assessment. A mobile platform or compute base should not be assumed to work merely because it is politically attractive. Any reduction in payroll contributions should be matched by scored revenue, not financed by an unfunded promise.

Finally, the state needs transition rights without a job-preservation fiction. Public contracts should require, where related to the contract, auditable evaluation, security records, portability, interoperability, exit plans, transition information and relevant skills provision. Human review should be defended for safety, due process and contestability, not because a verifier is expected to recreate the lost demand for many producers.

United Kingdom: first 24 months

The first UK task is to publish a baseline and a fiscal map. Government should ask for transparent modelling of household resources, employment, hours, earnings, rents, energy costs, market concentration, public-investment returns, employer National Insurance and full-expensing interactions. The OBR must score any permanent tax or welfare change before Parliament is asked to approve it. The published model should state uncertainty, behavioural assumptions, administrative costs and distributional effects.

Within that process, Parliament should consider focused welfare changes: a protected Agency Account, a revised capital treatment for that account, and smoother treatment of intermittent earnings. These are bounded-regret reforms because they help people facing ordinary insecure work as well as any future AI-driven displacement. They should not be sold as cost-free, nor as a full basic income by another name.

The National Wealth Fund should publish a separate public-interest return policy for relevant new investments. It should identify what counts as realised cash, what reserves must be retained, and what proportion could eventually flow into the Agency Fund. No existing £27.8bn capacity should be re-labelled as a household pot. A small, capped transition reserve may be established through the normal fiscal process, but it must be explicitly funded and time-limited.

Central government should pilot procurement clauses in suitable new digital and AI contracts. Clauses should be directly related to the contract and testable: data portability, interoperable exit, performance evaluation, supplier transparency, security, incident reporting and transition planning. The pilots should report bidder participation, administrative burden, compliance and value for money before wider use.

The UK does not need to wait for a general AI Act, because it has no such single framework. Sectoral regulators and procurement authorities can improve due process and market contestability now. Competition work should focus on bottlenecks and lock-in, while retaining independent case assessment and appeal rights. Housing and energy measures must be developed with local and devolved delivery bodies, because a centrally paid Agency Floor cannot itself build homes or reduce bills.

European Union and member states: first 24 months

EU institutions should not promise a permanent EU income entitlement. Existing adjustment and social funds can support retraining, inclusion and implementation capacity, but they do not constitute permanent income. Instead, EU action should create a common minimum measurement framework, model procurement clauses and coordinated competition priorities.

The common framework should distinguish adoption data from household outcomes. It should not mechanically compare UK business series with EU enterprise series, since the supplied measures cover different populations. It should standardise definitions, publication intervals, caveats and review methods while allowing national thresholds and fiscal decisions.

The EU AI Act should be implemented as it stands. Workplace uses such as recruitment and worker management can be high-risk, but the Act does not replace income. Proposed later high-risk deadlines were not finally enacted on 12 July 2026, so they should not be treated as operative law. Member states should use the Act for rights, redress and risk management, not as a disguised employment guarantee.

EU digital competition regimes can support portability, interoperability and data access. EU procurement guidance should make clear that contract conditions must be contract-linked, verifiable and non-discriminatory. They cannot become a route to unrelated parent-company equity, domestic preference or a general levy on suppliers.

The planned mobilisation of €200bn for AI investment, including €20bn for up to five gigafactories, should be used to improve public-interest conditions on relevant projects where legally permissible. Those conditions may include access, portability, evaluation and household-cost outcomes. They cannot be treated as an EU cash-transfer fund, and selective national support remains subject to State aid rules.

Member states should use the same first-clock package through their own systems: reform benefit cliffs where appropriate, build protected saving mechanisms, strengthen housing and energy policy, and decide whether national investment vehicles can generate public returns. National tax-and-transfer systems are the credible route for recurring income, with EU support for capacity and cross-border market rules.

Years 3 to 5 and dormant triggers

Years 3 to 5 should scale proven mechanisms, not assumptions. Procurement clauses should expand only where pilots show legal fit, usable compliance evidence, viable bidder participation and value for money. Competition remedies should be assessed for access, price, switching and enforcement outcomes. Agency Fund distributions should begin only from realised and independently audited receipts.

The Universal Agency Floor should have a two-key escalation rule. The following is a policy-design threshold, not a claim that it is presently met. After a published baseline has been established, the first key is breached when at least three of the following persist for six consecutive quarters against the published trend and seasonal method:

• The working-age employment rate is at least two percentage points below trend.

• Hours worked per working-age adult are at least 3% below trend.

• Median real weekly earnings are at least 5% below trend.

• Real household resources after housing and energy costs are at least 5% below trend.

The second key is independent confirmation that the data are not a short-lived measurement break or one-off shock, plus a fiscal certificate. A permanent rise in the Agency Floor requires projected recurring coverage of at least 110% of its cost over five years under published stress assumptions. A temporary supplement may be released from a pre-funded reserve, but it expires after 24 months unless renewed through the normal fiscal route.

The trigger deliberately does not ask whether AI caused a particular redundancy or accounts for a particular percentage of a workflow. If the outcomes worsen for another reason, temporary demand protection can still be justified. If AI adoption rises but the outcomes do not worsen, no trigger fires and no special tax is warranted.

A trigger should unlock support at national level. In the UK, Parliament retains the decision on permanent tax and welfare parameters. In the EU, member states make equivalent decisions through their systems. An EU-wide own resource or permanent central transfer should remain dormant unless the required unanimous political and legal route is actually secured.

Funding and fiscal arithmetic

There is not enough supplied arithmetic to claim that this package can presently fund mass agency at full scale. It cannot honestly do so. The UK’s public debt is close to 95% of GDP in the March 2026 forecast, and the National Wealth Fund’s £27.8bn capacity is not an annual dividend stream.

The scale is easy to state even where the national denominator is not supplied: every £100 per month paid to one million eligible adults costs £1.2bn a year before administration, interactions with other benefits or additional public services. A mass payment must multiply that figure by the number of eligible adults. No credible proposal should hide this multiplication.

The funding waterfall is therefore essential.

First, one-off land, planning, asset-sale or capital receipts should finance infrastructure, supply, capital reserves or a temporary adjustment reserve. They should not finance permanent entitlements.

Second, permanent agency payments require stable recurring receipts. Potential legal payers are the owners or recipients of properly defined domestic rents, taxable business surplus and public-investment cash returns. The economic burden may fall elsewhere, so modelling must test prices, wages, investment and avoidance rather than assuming that the legal payer bears the cost.

Third, public investment returns count only when cash has been realised. Paper valuation gains, projected savings and promotional investment totals do not count.

Fourth, borrowing should not finance a permanent Agency Floor. It may have a limited role in a separately authorised temporary stabilisation response, but debt close to 95% of GDP means that even this must be capped and justified.

Before enactment, governments must publish: eligibility numbers; payment levels; interactions with existing benefits and tax; administrative and fraud costs; revenue yield; tax incidence; behavioural responses; import treatment; offshore avoidance; the effect of employer National Insurance changes; the effect of full expensing; investment returns; debt and fiscal-rule implications; and regional housing and energy impacts. Until that arithmetic exists, the honest first-stage offer is lower essential costs, stronger savings buffers, portable rights and a limited funded reserve, not a fictitious fully funded dividend.

Political coalition and public case

The public case is simple: if technology makes valuable production cheaper, the gains should show up in household security and lower essential costs, not only in lower labour demand and higher rents. This is neither an attempt to stop useful technology nor a requirement that people pretend commercially unnecessary work is necessary.

Workers and unions gain an individual buffer, smoother transitions and enforceable due process. Firms gain policy neutrality about technology, clearer public-procurement requirements and a less brittle consumer base. Fiscal conservatives gain explicit funding gates, no automatic debt-financed permanent commitment and independent review. Local leaders gain a case for housing, energy and infrastructure that reduces cost pressure. Innovators gain access to public investment, but accept transparent public-interest terms when public capital bears risk.

The principal losers are likely to be recipients of protected scarcity rents, dominant platforms facing exit and interoperability duties, and some firms or asset owners facing a broader contribution base. The answer should be gradual phase-in, predictable rules, protection for viable investment and no retrospective confiscatory claim. Local retention of some planning-related value can also build support for supply.

Sequencing matters politically. Start with bills, savings buffers, fair procurement and portable markets. Build the delivery system before arguing over a large cash number. Make permanent expansion conditional on funding, not faith.

Durability and anti-capture design

The compact should divide power deliberately. Revenue collection, public investment, payment administration, competition enforcement and outcome evaluation should not sit in one opaque body. The Agency Fund needs separate ledgers, published accounts, independent audit, conflict rules and a public record of realised returns, reserves and distributions.

No UK statute can permanently bind a future Parliament. The practical protection against raids is visibility and procedural cost: a future government should need an explicit legislative vote, published fiscal certificate and statement of which account it is changing. The same transparency should apply to national funds in member states. A false promise of constitutional entrenchment would be less durable than an institution voters can see being diverted.

Anti-avoidance should attach to observable legal and economic connections, not vendor declarations about AI. Any new contribution base needs domestic nexus, treatment of imports and foreign suppliers, group-accounting rules, audit powers and review of relocation effects. International coordination is useful, but the package must not collapse if other countries adopt different rules.

Payment administration should minimise personal data and provide clear appeal rights. There should be no individual AI-exposure score, no requirement to surrender data in exchange for a payment and no central official deciding whether a person has been “replaced by AI”. Local institutions should shape housing and cost-reduction delivery within national standards, reducing both administrative centralisation and the risk of a single captured national allocator.

Legal and institutional obstacles

In the UK, Parliament can alter tax and welfare parameters, but major permanent transfers require a funding decision and OBR scoring. Universal Credit reform cannot be presumed costless. National Wealth Fund governance can be directed towards transparent public returns, but its existing mandate does not automatically create a citizen dividend.

UK procurement conditions must remain proportionate, objectively verifiable and tied to the contract. Competition enforcement must remain lawful and case-specific. The sectoral UK AI approach means the package should use existing regulatory and procurement routes rather than assume a new economy-wide AI employment regime.

At EU level, direct-tax decisions and new own resources normally require unanimity. The ordinary EU budget cannot run a deficit, and Article 123 TFEU blocks monetary financing by the ECB and national central banks. These are decisive reasons not to present a centrally financed permanent EU income scheme as an immediate answer.

EU procurement conditions must be contract-linked, verifiable and non-discriminatory. State aid rules constrain selective national support. Competition and portability can reduce lock-in but do not themselves distribute income. The AI Act can improve rights and risk control, but it is not a fiscal instrument. Member states remain indispensable, and their different fiscal space and political choices mean that a common EU dashboard should not become a uniform EU benefit rule.

Failure modes, review and exit rules

The first failure mode is false certainty. Adoption rates, self-reported headcount reduction and occupational exposure should never alone trigger a permanent payment or levy. If the outcome trigger is not met, the dormant income expansion remains dormant.

The second is rent capture. If median housing or energy costs rise faster than post-transfer household resources for four consecutive quarters after an Agency Floor increase, the next increase pauses. New funding should be redirected to cost-reduction and supply measures until an independent review explains the pass-through.

The third is fiscal illusion. If realised fund returns, defined contribution receipts or compliance revenue fall below forecast, the recurring payment cannot be maintained by drawing down capital or by automatic borrowing. Payment increases pause, and any temporary supplement expires on schedule.

The fourth is administrative overreach. Procurement clauses that reduce bidder participation, fail contract-link tests or create unmanageable reporting should be narrowed or ended after evaluation. Competition remedies that do not improve switching, access or prices should be revised through ordinary legal processes, not used as proof of a broader economic thesis.

The fifth is treating a conventional recession as confirmation of technological displacement. The outcome trigger can still release a temporary reserve, but permanent institutional change requires the fiscal certificate and a review of alternative explanations. It does not need a speculative causal attribution to AI.

Reviews should occur at 24 months and 60 months. The 24-month review tests data quality, delivery readiness, procurement pilots and fiscal modelling. The 60-month review decides whether the special trigger apparatus remains useful. If adoption continues without persistent adverse household outcomes, the temporary escalation mechanism should lapse unless renewed. If the data cannot support reliable outcome measurement, it should be simplified or ended rather than turned into a discretionary black box. A recovered transition supplement should taper out only after the outcome thresholds have been clear for eight quarters; a separately funded core entitlement should not be casually cut.

Feasibility table

High means the relevant level can begin through a policy function it already performs. Medium means primary legislation, technical modelling, coordination or sustained political agreement is needed. Low means treaty, budget or unanimity constraints, or an unproven recurring revenue base, are central blockers.

What is genuinely new here

The distinctive feature is not the existence of transfers, public investment, competition policy or welfare reform. It is their ordering and accounting.

First, the package creates a rent-to-agency waterfall: one-off gains fund supply and capital; measurable cost reductions improve real living standards; only recurring and realised receipts fund recurring cash. This prevents a common error in which an attractive public asset is counted three times, as investment, savings and dividend income.

Second, it uses an outcome-and-funding escalator rather than an adoption trigger. It is deliberately agnostic about whether an adverse labour outcome is attributable to AI at the individual workflow level. That makes it administrable and useful even if the thesis is partly wrong.

Third, it treats a protected individual savings buffer as a constitutional component of agency, not merely a welfare parameter. That is a modest but important answer to a system in which a person may need to move repeatedly between paid work, care, training and periods of low market income.

Fourth, it confines public-return claims to investment arrangements where public capital actually bears risk. It does not misuse procurement as a route to seize parent-company equity or national preference.

Bottom line

Prepare a successor circuit before declaring the old one broken. Start by lowering fixed costs, protecting savings, keeping markets contestable and building state capability. Scale cash support only when sustained household outcomes justify it and recurring revenues truly cover it. That is compatible with AI adoption, honest about fiscal limits and valuable even if Unit Cost Dominance never becomes economy-wide.