The Agency Floor: Preserving Mass Economic Participation Through the AI Transition

Doctrine in one sentence

Tax observable business revenue and capital deployment, not unmeasurable AI usage, to fund portable individual agency accounts that function whether the wage circuit holds, weakens, or breaks.

Executive summary

The Unit Cost Dominance thesis identifies a conditional but serious risk: that AI-driven productivity gains could progressively decouple business revenue from mass wage income, undermining both household demand and democratic legitimacy. A credible response cannot rely on slowing adoption, measuring AI's share of each task, or coordinating restraint across jurisdictions. It must instead build successor institutions that channel economic agency to individuals regardless of whether their labour remains commercially necessary.

This brief proposes a seven-plank package centred on three principles. First, shift the marginal tax base from labour towards value added and capital, using instruments that governments already know how to administer. Second, create portable, individually controlled Agency Accounts that provide durable purchasing power without the dependency traps of means-tested welfare. Third, use procurement, competition law and housing policy to prevent rents from absorbing the new income before households can use it.

Every plank is designed for bounded regret. If the thesis proves substantially wrong and wage employment remains robust, the package still corrects the existing tax bias favouring capital over labour, strengthens household resilience, improves competition in digital markets, and moderates housing costs. Nothing in it requires job preservation mandates, AI usage measurement, or international coordination as a precondition.

The policy package

The package contains seven planks, described briefly here and elaborated under jurisdictional headings below.

Plank 1: Employer Value-Added Contribution (EVAC). A broad-based levy on business value added (revenue minus purchased inputs), replacing part of employer payroll contributions. Because it falls on value added regardless of input mix, it does not penalise firms for hiring humans or reward them for replacing workers with AI. It is observable, auditable through existing VAT infrastructure, and familiar in principle from France's former taxe professionnelle and current cotisation sur la valeur ajoutée des entreprises (CVAE).

Plank 2: Individual Agency Accounts. Personal accounts, owned by residents, receiving regular credits funded by EVAC and other plank revenues. Withdrawals are unrestricted above a modest floor balance. Unlike Universal Credit, they are not means-tested against savings and do not create poverty traps. They resemble a hybrid of auto-enrolment pensions and Alaska's Permanent Fund Dividend, but liquid and lifelong.

Plank 3: Capital Expensing Rebalance. The UK's permanent full expensing currently allows 100% first-year deduction for qualifying plant and machinery. This plank does not remove full expensing but subjects firms claiming it above a threshold to a modest top-up EVAC rate, linking the capital subsidy to the broader revenue base. It ensures that accelerated capital deployment contributes proportionally to the agency floor.

Plank 4: Procurement Transparency and Portability Standards. Public contracts above a threshold require suppliers to publish workforce transition plans, use interoperable data formats, and offer portability of any AI-generated institutional knowledge to successor contractors. This is contract-linked, verifiable and non-discriminatory, and therefore compatible with existing EU procurement law.

Plank 5: Digital Competition and Rent Prevention. Strengthen data portability and interoperability obligations under the UK Digital Markets, Competition and Consumers Act and the EU Digital Markets Act. The goal is to prevent platform lock-in from allowing a small number of AI intermediaries to extract rents that absorb agency-account income. Competition enforcement does not distribute income, but it defends the purchasing power of income that is distributed.

Plank 6: Housing Cost Anchor. Any unconditional transfer to households risks capitalisation into housing costs, especially in supply-constrained markets. This plank links agency-account funding to parallel obligations on local and national planning to release housing supply. In the UK, this means reinforcing and extending existing planning reform. In the EU, it means member-state commitments under National Reform Programmes.

Plank 7: Contingent Compute Levy. A dormant levy on large-scale compute procurement (measured in purchased or leased compute capacity above a high threshold), activated only if observable labour-market indicators cross defined triggers. This provides a fiscal backstop without immediate cost to the AI sector and without requiring measurement of AI usage within firms.

United Kingdom: first 24 months

Months 1 to 6: Legislative and institutional preparation.

• HM Treasury commissions the OBR to model EVAC at illustrative rates (0.5%, 1.0%, 1.5% of value added) with corresponding reductions in employer National Insurance. The modelling must estimate revenue, distributional effects and behavioural responses. No enactment occurs without this scoring.
• HMRC begins a feasibility study for EVAC collection, assessing whether existing VAT return infrastructure can serve as the administrative backbone. Value added for EVAC purposes would be defined as turnover minus VAT-deductible purchases, a figure firms already report.
• The Department for Work and Pensions designs the Agency Account framework, building on existing auto-enrolment and NS&I infrastructure. Key design choices: universal residency-based eligibility (replacing part of Universal Credit for those who receive both), individual ownership, no means test on accumulated balance, annual credit rather than monthly wage replacement in the first phase.
• DSIT publishes a formal set of labour-market monitoring indicators (see Failure Modes below) and commits to biannual review.

Months 7 to 18: Pilot and legislative passage.

• A Finance Bill introduces EVAC at the lowest modelled rate, initially as a supplement rather than a replacement for employer NI, to allow parallel running and revenue comparison. The OBR scores this as a permanent measure.
• The Agency Account pilot launches in two or three local authority areas, offering a modest annual credit (illustratively £500 to £1,000 per adult) to all residents, funded from EVAC pilot revenue and NWF transitional support. The NWF's digital and technology remit provides legal cover for transitional capitalisation.
• The Digital Markets, Competition and Consumers Act is used to designate major AI platform providers with strategic market status, imposing interoperability and data portability obligations.
• Planning reform measures already in progress are explicitly linked to the Agency Account programme: local authorities that fail to meet housing delivery targets face conditional reductions in agency-account funding allocations (creating political pressure for supply release rather than punishing residents directly).

Months 19 to 24: Evaluation and adjustment.

• The pilot is independently evaluated. Key questions: did recipients spend or save the credit? Did local housing costs absorb it? Did administrative costs exceed projections?
• Parliament debates the first EVAC rate adjustment and the national rollout timetable for Agency Accounts.

European Union and member states: first 24 months

The EU's institutional constraints are tighter. Direct taxation requires unanimity. The EU budget cannot run a deficit. Agency Accounts must therefore be a member-state instrument, with the EU providing coordination, standards and complementary action.

EU-level actions (European Commission and Council):

• The Commission publishes a Recommendation on Employer Value-Added Contributions, providing a model EVAC design and inviting member states to adopt compatible frameworks. This has no binding force but creates a coordination focal point and reduces first-mover disadvantage.
• The Commission proposes an amendment to the European Social Fund Plus regulation, allowing ESF+ funds to co-finance the administrative setup costs of national Agency Account systems in member states that adopt them. This uses existing legal bases and budget lines, not new own resources.
• The AI Act's high-risk classification for workplace AI systems is operationalised through implementing acts. The Commission ensures that transparency and record-keeping obligations under the AI Act generate data useful for labour-market monitoring, without creating a separate AI-measurement regime.
• The Digital Markets Act's existing interoperability and portability provisions are enforced with priority against AI-intermediary gatekeepers.
• Procurement: the Commission updates public procurement guidance to include model clauses for workforce transition transparency and data portability in AI-related contracts.

Member-state actions (leading coalition of willing states):

• France, Germany and two or three willing member states pilot EVAC designs, drawing on France's existing CVAE experience. France is the natural first mover because it already operates a value-added business tax and has the administrative infrastructure.
• These states design national Agency Accounts, initially funded through partial EVAC revenue and reductions in employer social contributions.
• Member states link Agency Account rollout to National Reform Programme housing commitments, providing the rent-prevention anchor.

Years 3 to 5 and dormant triggers

Year 3: UK national rollout of Agency Accounts if the pilot evaluation is positive. EVAC rate rises to the medium modelled rate, with a corresponding reduction in employer NI. The net effect on total employer contribution rates should be approximately neutral at this stage; the shift is in the base, not the burden.

Year 4: The EU Commission reviews member-state adoption and, if a critical mass exists, proposes a Directive on minimum standards for value-added business contributions, using the social-policy legal base (Article 153 TFEU, qualified majority for working conditions). This is legally contested territory and may require Treaty-base litigation, but it is the most plausible route to binding coordination without unanimity on tax.

Year 5: First review of dormant triggers.

Dormant triggers and the Contingent Compute Levy:

The Compute Levy activates only if two of the following three indicators are sustained for four consecutive quarters:

1. The employment-to-population ratio for 25-to-54-year-olds falls more than 3 percentage points below its 2025 baseline in the relevant jurisdiction.
2. The labour share of national income (compensation of employees as a share of GDP) falls more than 4 percentage points below its 2025 level.
3. Median real household income falls for three consecutive years despite positive real GDP growth.

If triggered, the levy applies to purchases or leases of compute capacity above a high threshold (initially set to exclude all but the largest deployments). Revenue flows to Agency Accounts. The levy is designed with import parity: compute purchased abroad for domestic use is included; compute sold abroad is excluded. This prevents simple offshoring of the base.

If the indicators recover, the levy is suspended after two consecutive quarters above threshold. It is not permanent by design.

Funding and fiscal arithmetic

Honest gaps first. The precise revenue from EVAC at any given rate depends on behavioural modelling that has not been done. The following are illustrative and must be scored by the OBR (UK) and member-state fiscal councils (EU) before enactment.

UK illustrative figures:

• UK gross value added (GVA) was approximately £2.1 trillion in 2025. A 1% EVAC on all business value added would yield roughly £21 billion gross before behavioural responses, exemptions for small businesses, and administrative costs. A small-business exemption (firms below the VAT threshold) would reduce the base significantly, perhaps to £16 to £18 billion.
• Employer NI currently raises approximately £100 billion. A 1% EVAC could therefore fund a partial reduction in the employer NI rate (perhaps 1 to 2 percentage points), making the shift roughly revenue-neutral in the near term. Agency Account credits would initially be funded from the incremental revenue above the NI offset, plus transitional NWF support.
• At £1,000 per adult (approximately 52 million adults in the UK), a universal Agency Account credit costs approximately £52 billion per year. This is not fundable from a 1% EVAC alone. The initial credit must therefore be modest (£500 or less), scaled up only as EVAC rates rise and employer NI rates fall.
• The arithmetic gap is real and must not be hidden. A universal credit large enough to provide meaningful agency (perhaps £3,000 to £5,000 per year) requires either a substantially higher EVAC rate, activation of the Compute Levy, or reallocation from other spending. The package is designed to scale, not to promise full funding from day one.

EU: Member-state fiscal arithmetic varies enormously. France's CVAE raised approximately €9 billion before its recent reduction. The principle is the same: shift the base, start modestly, scale conditionally.

Transition funding: The UK NWF provides bridging capital for the pilot phase (within its existing £27.8 billion capacity). EU member states can draw on ESF+ for administrative setup. Neither source is sufficient for permanent funding; they buy time for EVAC to mature.

Political coalition and public case

UK coalition: The package appeals to a centre-left government seeking to modernise the tax base and strengthen household resilience, and to business voices frustrated by employer NI increases that penalise hiring. The EVAC-for-NI swap is presentable as "backing businesses that hire" while ensuring that businesses which automate also contribute. Conservative supporters of full expensing can accept the rebalance if the net employer burden is broadly neutral. Trade unions gain a new institution (Agency Accounts) that supports members facing displacement.

EU coalition: France is the natural champion given its CVAE history. Germany's coalition may support a coordinated approach that prevents regulatory arbitrage. Nordic states with strong social models may see Agency Accounts as a natural extension. Opposition is likely from low-tax member states (Ireland, Luxembourg) who fear base erosion, and from states wary of any EU coordination on taxation.

Public case in ordinary language: "As technology changes the economy, everyone should share in the gains. We are shifting part of business contributions from a tax on hiring to a tax on business success, so that companies contribute fairly whether they grow through people or through technology. Every adult receives a personal Agency Account, money that is yours, not means-tested, not dependent on your employer. This is not a handout. It is your share of the productivity our economy creates."

Losers and compensation: Capital-light, high-value-added firms (software, finance, consultancy) face a higher relative burden under EVAC than under payroll taxes alone. This is partially offset by reduced employer NI. Firms with very high value added per worker are the clearest losers. Compensation comes through the NI reduction and through the demand stimulus of Agency Account spending.

Durability and anti-capture design

• Fiscal raids: Agency Accounts are legally structured as individual property, not government spending. Accumulated balances cannot be redirected to general revenue without primary legislation, creating a political cost for raids.
• Dilution: The EVAC rate is set by Parliament or member-state legislature, not by ministerial discretion. Rate reductions require the same legislative process as rate increases.
• Avoidance and offshoring: EVAC is based on value added generated in the jurisdiction (domestic turnover minus domestic purchased inputs), using the same territorial principle as VAT. Transfer-pricing risks exist but are no worse than for existing VAT and corporation tax. The Compute Levy includes import parity.
• Platform capture: The Agency Account is a government-administered financial account (like NS&I), not a platform product. No private intermediary controls access or extracts fees.
• Hostile future governments: The account structure creates a constituency of recipients (every adult) who will defend it, similar to the political durability of state pensions and the NHS. Universality is the strongest defence against erosion.
• Concentration of administrative power: HMRC already administers PAYE, NI and VAT for tens of millions of accounts. Agency Accounts add a disbursement function, not a surveillance function. No new regulator is created.

Legal and institutional obstacles

UK: EVAC requires primary legislation (Finance Act). Agency Accounts require primary legislation establishing the account framework and its relationship to Universal Credit. Neither faces a constitutional barrier. The main obstacle is OBR scoring: if the OBR judges EVAC as contractionary (because it taxes a broader base than NI), the fiscal headroom for Agency Account credits shrinks. This is a political constraint, not a legal one.

EU: A binding Directive on value-added business contributions faces Treaty-base disputes. Article 153 TFEU permits qualified-majority legislation on working conditions, but the Court of Justice may rule that a business-contribution floor is a tax measure requiring unanimity under Article 115. The Commission Recommendation route avoids this but has no binding force. Enhanced cooperation (minimum nine member states) is a fallback but risks fragmentation.

Procurement: EU procurement conditions must relate to the subject matter of the contract and be verifiable. Workforce transition transparency is defensible; requiring equity stakes in parent companies is not. The package stays within this boundary.

State aid: Member-state EVAC designs must not function as disguised subsidies. A uniform rate with a small-business exemption is unlikely to raise State aid concerns. Sector-specific exemptions would.

Failure modes, review and exit rules

Failure mode 1: EVAC depresses investment. If business investment falls significantly (more than 5% below trend) in the two years after EVAC introduction, the rate is frozen and the OBR conducts a special review. If investment does not recover within a further year, the EVAC rate is reduced to the pilot level.

Failure mode 2: Agency Account credits are absorbed by housing costs. The biannual review includes a housing-cost pass-through analysis. If median rents in Agency Account pilot areas rise faster than in control areas by more than 2 percentage points, housing supply obligations are escalated before the credit is increased.

Failure mode 3: The thesis is substantially wrong. If employment-to-population ratios, labour income shares and median real incomes all remain stable or improve over a five-year period, the package remains in place as a tax-base modernisation and household-resilience measure, but the Compute Levy remains dormant and Agency Account credits are not escalated beyond modest levels. The bounded-regret value is a fairer tax base and a small universal savings instrument.

Failure mode 4: AI adoption is faster than expected. If the dormant-trigger indicators are breached within the first 24 months, the Compute Levy activation is brought forward. Agency Account credits are scaled up using Compute Levy revenue. This scenario validates the package design rather than breaking it.

Review cycle: Biannual review by an independent body (a standing commission or the OBR with an expanded mandate) reporting to Parliament. The review publishes indicator dashboards, evaluates pass-through effects, and recommends rate adjustments.

Exit rule: Any plank can be suspended or repealed through the normal legislative process. The package does not create irreversible commitments. Agency Account balances, once credited, belong to individuals and are not clawed back on programme exit.

Feasibility table

Arithmetic is missing for: precise EVAC revenue after behavioural responses and small-business exemptions; Agency Account credit levels that are both fundable and meaningful; Compute Levy revenue at various thresholds; and housing supply elasticity in response to planning reform linkage. All require formal modelling before enactment.

What is genuinely new here

Most AI policy discussion focuses on retraining, job guarantees, universal basic income, or robot taxes. This package departs in several ways.

The tax-base shift, not the tax level, is the core move. EVAC does not increase the total employer contribution burden in the near term. It changes what is taxed: value added rather than headcount. This removes the perverse incentive in payroll taxes that makes hiring more expensive relative to automation. It is not a robot tax (which requires identifying robots) or an AI levy (which requires measuring AI use). It taxes the same thing VAT taxes, but as an employer contribution rather than a consumption tax, preserving the link between business activity and social funding.

Agency Accounts are not UBI. They are individually owned, accumulating, and unrestricted. They do not replace wages or pretend to. They provide a floor of economic agency that persists whether someone is employed, self-employed, caring, studying, or displaced. The accumulation feature and absence of means-testing distinguish them from both Universal Credit and from flat UBI proposals. They create a constituency with a visible, personal stake, which is the strongest political defence against erosion.

The dormant-trigger architecture is genuinely contingent. The Compute Levy does not exist until observable labour-market indicators breach defined thresholds. This means the package imposes no cost on the AI sector unless and until the thesis is validated by real-world outcomes. It provides fiscal insurance without speculative taxation.

The housing anchor is integrated, not afterthought. Most transfer proposals ignore the likelihood that unconditional income will be captured by landlords in supply-constrained housing markets. Linking Agency Account rollout to housing supply obligations is unusual and politically difficult, but without it the transfer is a landlord subsidy.

Bottom line

The wage-demand circuit may hold, weaken gradually, or fracture. This package does not bet on one outcome. It shifts the tax base from labour towards value added (useful regardless), creates a portable, individually owned agency floor (useful regardless), enforces digital competition (useful regardless), and holds a fiscal backstop in reserve (costless unless needed). The arithmetic is incomplete and must be modelled before enactment. The first 24 months require only a pilot-scale commitment and no irreversible fiscal exposure. The honest gap is that a universal credit large enough to replace significant lost wage income requires either a much higher EVAC rate or activation of the Compute Levy, and the political appetite for either is untested. The package is designed to scale into that gap if reality demands it, and to remain a modest but valuable reform if it does not.