Foundations for Agency: Reducing Rents, Widening Access and Embedding Adjustment Capacity

Doctrine in one sentence

Maintain mass economic agency and demand by constraining non-labour costs and rents, widening access to verifiable skills and small capital, and embedding observable adjustment capacity in existing tax, welfare, procurement and competition systems, without requiring measurement of AI intensity or coordination on adoption.

Executive summary

The Unit Cost Dominance thesis identifies a contingency. If AI plus verification delivers broad cognitive work at lower unit cost than human-only production, mass wage labour may cease to be the main route to independent economic agency. The mechanism starts in digitally mediated cognitive work and spreads through interface collapse and coordination incentives. It does not predict the disappearance of all jobs or set a timetable. June 2026 ONS data show 29 per cent of UK businesses using AI, rising to 49 per cent for firms with 250 or more employees, with only 5 per cent of AI-using businesses reporting headcount reduction. In the EU, 20 per cent of enterprises with at least 10 employees used AI in 2025. Evidence on hiring effects remains provisional.

A response must work if adoption continues. It must also deliver ordinary public benefits if the thesis proves largely incorrect. It cannot depend on preserving commercially unnecessary jobs, on precise measurement of the AI share of workflows, or on simultaneous restraint by all countries.

The package contains seven planks in three pillars. The first lowers non-wage costs through competition enforcement and public contract terms. The second widens access to small capital and earnings by reforming means tests and directing public investment returns toward households. The third embeds observable triggers and review points in existing funds, procurement and statistics so support can scale or contract on published data.

Jurisdictional fit is respected. The United Kingdom can act on welfare, tax, procurement and the National Wealth Fund. The European Union can act through competition law, procurement directives, the AI Act and coordination of existing social funds. Member states handle most recurring transfers.

Fiscal credibility is bounded. The package uses the National Wealth Fund's £27.8 billion capacity, the EU's planned €200 billion AI mobilisation, and reallocation within existing funds. It proposes modelling of tax changes rather than new levies. It creates no large permanent entitlements. Where arithmetic is missing, this is stated. If displacement proves extensive, these measures are a foundation, not a full replacement for the wage-demand circuit.

Distribution works through lower platform and digital service prices, reduced welfare cliffs, contract-driven skills formation, and any returns from public capital. Durability rests on published statistical triggers, dispersed administration and explicit review points. The design is adoption-compatible and falsifiable: high adoption without displacement pressure allows scaling back.

The policy package

Seven planks attach to observable administrative bases and existing or readily exercisable powers.

Plank 1 reforms asset and earnings rules in means-tested support. For the United Kingdom this means raising the Universal Credit capital limit above £16,000 and expanding earnings disregards. The base is existing benefit administration. The aim is to let households hold small buffers and combine modest earnings with support. Analogous reforms are available to member states.

Plank 2 introduces contract-linked conditions in public procurement above defined thresholds. Contracts would require disclosure of intended AI use, documentation of human oversight for high-risk systems under the AI Act or UK sectoral guidance, skills and transition clauses, and reasonable data or output portability. Conditions must be relevant to the contract, verifiable and non-discriminatory.

Plank 3 directs competition authorities to prioritise remedies that reduce lock-in and rent extraction in digital layers supporting AI services. This includes data portability, interoperability and access remedies in cloud, model distribution and related services. Lower rents raise household and small-business purchasing power.

Plank 4 creates triggered deployment rules for existing social, retraining and regional adjustment funds. Published indicators from ONS, DSIT and Eurostat would serve as reference points for faster or larger drawdown. Indicators could cover adoption rates, changes in hiring or hours in exposed occupations, and regional shifts. No new permanent EU-level income entitlement is created. Member states remain the route for recurring support.

Plank 5 requires governance changes to public AI investment vehicles so that a defined share of returns serves household-accessible benefits. For the National Wealth Fund, with its £27.8 billion capacity and digital remit, this means investment criteria favouring measurable cost reductions or capability gains for households, plus design work on recycling a portion of equity returns. Equivalent steps would apply to member-state vehicles and the EU €200 billion mobilisation. Existing mandates do not create citizen dividends; new rules and modelling are required.

Plank 6 commissions a review of tax parameter interactions affecting the relative cost of labour and capital in cognitive tasks. In the UK this covers the 15 per cent employer National Insurance rate and permanent full expensing for qualifying plant and machinery. The review would model revenue, behavioural and incidence effects. Any changes would proceed through normal budget processes. EU scope is limited by unanimity requirements on direct tax.

Plank 7 shapes implementation of high-risk obligations under the EU AI Act and UK sectoral practice. In recruitment and worker management, transparency, oversight and documentation requirements would preserve points of human judgement and review. UK sectoral regulators would be encouraged to adopt comparable expectations. The plank maintains institutional space for human involvement in verification without assuming this restores the full wage circuit.

The planks work together. Procurement and competition constrain rent capture. Welfare reform allows small capital. Investment vehicles and funds support capability. Tax review addresses incentives. Triggers and high-risk rules provide observability.

United Kingdom: first 24 months

The United Kingdom can move fastest on welfare, procurement and National Wealth Fund governance.

In the first six months the Department for Work and Pensions and HM Treasury would publish modelling of options for raising the Universal Credit capital limit and adjusting earnings disregards and tapers. Any legislation would follow Office for Budget Responsibility scoring.

Cabinet Office would issue updated procurement policy notes requiring AI disclosure, oversight documentation and skills clauses. Three departments would pilot the conditions. The National Wealth Fund would receive guidance directing part of its digital allocation toward projects with household cost or capability benefits and requiring publication of options for return-recycling mechanisms.

The Competition and Markets Authority would identify two to three priority cases or studies in layers supporting generative AI services, with explicit consideration of interoperability and data access remedies. The Office for National Statistics and Department for Science, Innovation and Technology would expand the business survey and begin a regular "Agency Indicators" series on adoption, employment and cost data.

Sectoral regulators would receive direction to align oversight expectations with high-risk themes from the AI Act. Devolved administrations would be invited to participate in indicator design and pilots. No new primary AI legislation is required.

European Union and member states: first 24 months

EU-level action centres on guidance, coordination and implementation of existing instruments.

The European Commission would publish interpretative guidance and model clauses on how procurement directives can accommodate AI-related conditions on transparency, oversight, skills and portability while remaining linked to the contract subject matter.

DG Competition, with national authorities, would issue a priority list of investigations and remedies focused on data access and interoperability in AI-enabling services, using existing tools.

The AI Act is already in force. Member states would complete designation of national authorities. The Commission would monitor implementation with attention to high-risk employment uses and issue guidance preserving human review points.

Existing social and adjustment funds would be front-loaded in regions with high adoption per the 2025 enterprise survey. This requires no new own resources. The Commission would coordinate common indicator sets.

Direct tax and new own resources remain subject to unanimity and are not pursued. Member states would begin parallel modelling of welfare and tax parameters. National investment vehicles would receive recommendations to align digital allocations with household-return criteria.

Years 3 to 5 and dormant triggers

Procurement conditions would extend to wider contract categories once pilots show feasibility. Competition remedies would be monitored for compliance and price effects.

If published ONS and European data show sustained divergence between AI adoption and employment or hours in exposed occupations exceeding pre-defined thresholds, triggered deployment of adjustment funds would increase. Thresholds would be conservative and published, for example a doubling of decline rates relative to baseline for four quarters. Activation would use existing mechanisms.

National Wealth Fund and parallel vehicles would move from design to operation of return-recycling mechanisms once cash flows appear. Early mechanisms would be modest and subject to fresh scoring.

The tax review would produce options. Legislatures would decide. Changes would be modest and reversible.

Dormant provisions would allow further design on broader asset instruments only if adoption and displacement indicators breach higher levels and fiscal headroom or new revenue has been identified. Activation would require formal decision after independent statistical review.

Funding and fiscal arithmetic

The package draws on three categories of resource.

First, reallocation within known investment envelopes. The National Wealth Fund has £27.8 billion capacity. The EU plans €200 billion for AI investment, including €20 billion for up to five gigafactories. Directing a share toward household-benefit projects requires governance change, not new money. Returns from equity positions could be partially recycled under new rules. Precise proportions and yields require modelling that does not currently exist.

Second, parameter changes within existing tax and welfare systems. Employer National Insurance and full expensing interact in the UK. Universal Credit capital and earnings rules are adjustable. The static cost of raising the capital limit depends on claimant distributions near the threshold and behavioural response. Current caseloads and asset data must be modelled by the Department for Work and Pensions and the Office for Budget Responsibility before enactment. Any net cost would need offset or acceptance.

Third, compliance and administrative effort. Procurement conditions impose costs on contractors. Competition remedies impose costs on firms. Adjustment funds are already funded; the plank affects timing and scale of drawdown.

No general AI, compute or robot levy is proposed. Any such instrument would require an observable base, import treatment, anti-avoidance rules and integration with existing taxes. None of these has been designed.

The package cannot fund large-scale recurring income replacement. At debt close to 95 per cent of GDP, and with the ordinary EU budget unable to run a deficit, major new permanent transfers would require sustained higher growth, reallocation or new revenue. The design therefore emphasises cost reduction, which multiplies household income value, and small-scale asset access. If unit cost dominance materialises at scale, this is a necessary but not sufficient component of a larger response.

Political coalition and public case

A viable coalition spans fiscal conservatives who value bounded commitments and triggers, pro-competition liberals who welcome lower digital rents, trade unions and centre-left parties seeking worker participation and skills protections, and regional representatives. Business groups favouring clear rules and skills pipelines can join if conditions remain proportionate.

The ordinary-language case is straightforward. Households face rising platform costs in services that mediate work and consumption. Welfare rules penalise small savings. Public contracts can require suppliers to develop human capabilities. Public AI investment can be structured so some returns reach citizens. These steps make sense whether adoption stays modest or accelerates. They require no agreement on exact pace or on coordinated restraint.

Losers are concentrated: large platforms facing interoperability obligations, some contractors bearing compliance costs, and taxpayers if welfare changes are not fully offset. Gains are diffuse: lower service prices, greater household buffers, and a more gradual transition.

Sequencing favours early low-visibility steps on procurement guidance and statistics. Welfare reform requires budget packaging. Competition cases take time. Investment changes can use existing mandates.

Durability and anti-capture design

Triggers rest on published official statistics, not self-reporting or ministerial discretion. Administration is dispersed across welfare departments, contracting authorities, competition bodies and investment vehicles. No single agency controls the package.

Procurement conditions must be published and challengeable. Competition remedies are appealable. Tax changes follow normal parliamentary process. Offshoring is constrained where the base is domestic. Procurement conditions can require oversight activities tied to contract delivery if non-discriminatory. Competition enforcement applies to services supplied into the jurisdiction.

Hostile future governments face visible reversal costs. EU procurement and competition rules are embedded in treaty competences. AI Act obligations are already in force. Capture is addressed by mandatory publication of investment criteria, templates and methodologies, plus regular public dashboards.

Legal and institutional obstacles

In the United Kingdom the obstacles are fiscal and political. Parliament controls tax and welfare. Procurement policy can be updated through guidance and secondary legislation. The National Wealth Fund can receive supplementary directions. The Competition and Markets Authority has digital competition powers. The main constraints are Office for Budget Responsibility scoring and the legislative timetable.

At EU level, direct tax and new own resources require unanimity. The ordinary budget cannot run a deficit. Article 123 TFEU blocks monetary financing. These rule out large new EU transfers. Within existing competences, obstacles are lower. Competition, procurement directives and the AI Act provide operable bases. Social funds can be redeployed without new legislation. Member states can move at different speeds on welfare and national vehicles.

Procurement conditions must be linked to the contract subject matter, verifiable and non-discriminatory. This limits use for unrelated equity transfer or domestic preference. Institutional capacity varies across member states.

Failure modes, review and exit rules

Procurement conditions may be too loose to bind or too tight and deter bidders. Competition cases may produce weak remedies after delay. Welfare changes may increase costs without asset gains if other prices absorb the margin. Investment mechanisms may yield little cash flow. Triggers may fire on noisy data or miss real but diffuse displacement.

Review points are built in. An annual published report would combine ONS, DSIT, Eurostat and national data on adoption, employment in exposed categories, wage and hours distributions, digital service costs, and welfare caseloads near thresholds. Formal reviews would occur at 24 and 60 months by a panel with statistical and fiscal expertise.

Decision rules would be pre-specified. If adoption remains below 40 per cent after five years and there is no statistically significant acceleration in hiring decline in exposed occupations, the presumption would be to scale back procurement stringency and fund triggers. If digital service cost indices fall after competition interventions, remedies would be maintained. If welfare net costs exceed modelled amounts without measurable asset accumulation, parameters would be reviewed for reversal.

Exit provisions would be explicit. Procurement conditions could be suspended by order following review. Welfare changes can be reversed by the same process used to enact them. Investment guidance can be rewritten. Statistical series would continue.

The package is falsifiable. If adoption proceeds rapidly, displacement appears in data, and household agency metrics do not deteriorate, the premise that non-wage channels are insufficient would be weakened. If adoption proceeds and agency metrics deteriorate while the package operates, the premise that it is adequate at scale would be weakened.

Feasibility table

The table uses High, Medium or Low only after the reasoning set out in the columns and surrounding text. Arithmetic remains missing for the precise static and dynamic costs of welfare parameter changes, the expected yield from National Wealth Fund return mechanisms, and the price effects of specific competition remedies. These gaps must be filled before enactment or scaling.

What is genuinely new here

The package does not restate the diagnosis. It proposes specific mechanisms.

It treats cost reduction and rent constraint as first-order contributions to demand and agency. Competition and procurement shape the terms of adoption rather than blocking it.

It designs small-scale asset access through reform of existing means tests and governance of public investment vehicles, rather than new capital grants or universal basic income. The approach is incremental and reversible.

It embeds falsifiability and scaling rules tied to official statistics from the outset.

It allocates roles by actual competence. The United Kingdom moves on welfare and the National Wealth Fund. The European Union moves on competition, procurement guidance and AI Act implementation. Member states move on their own transfer systems. No assumption is made that unanimity will be achieved on tax.

It acknowledges the gap between current fiscal envelopes and hypothetical requirements of mass agency replacement, rather than presenting the package as complete. It states where modelling is absent and where measures would be insufficient if displacement proves extensive.

It integrates the EU AI Act's high-risk provisions as a partial source of friction that can be shaped to preserve review points, without claiming oversight will restore the postwar wage circuit.

Bottom line

The United Kingdom and the European Union can take concrete, observable steps within 24 months that improve household agency and economic resilience whether or not unit cost dominance materialises across cognitive work. The measures use existing powers, attach to administrable bases, and contain explicit review and exit provisions. They do not depend on measuring the AI share of workflows or on every jurisdiction moving in lockstep.

They are not sufficient, on current fiscal parameters, to preserve mass economic agency if a broad share of cognitive labour loses unit-cost competitiveness and no other channel replaces the wage-demand circuit at scale. In that contingency the package would function as cost shields, capability platforms and adjustment mechanisms that reduce the speed and disorder of transition while additional revenue or spending choices are debated. That limitation is stated rather than obscured.

The design meets the tests of adoption compatibility, observable base, jurisdictional fit, bounded regret and falsifiability. Its political and institutional durability will depend on transparent execution, honest scoring and the willingness of future governments to adjust course on the basis of the data the package requires to be published.