Require human review of workplace AI: calculation note

Central fiscal result

+£1.0bn - Net public-finance impact in 2027-28

Low case: +£0.2bn. High case: +£4.0bn. Positive numbers are fiscal costs or borrowing pressure. Negative numbers are Exchequer savings or receipts.

Scenario and baseline

• Require assessments, worker consultation and human review for AI hiring, monitoring and discipline.
• Baseline is current law and published official data unless stated.
• Private business costs are excluded unless they affect tax or procurement.
• Target year is 2027-28, with later years shown separately.

Affected population

• Unit is employers and AI-covered decisions.
• No official affected-count estimate exists.
• Public-sector systems are directly fiscal.
• Private compliance costs are mostly off-budget.

Gross impact

• Central public-sector compliance cost is £0.60bn.
• Regulator and enforcement capacity adds £0.25bn.
• Public procurement/audit duties add £0.20bn.
• No fiscal value is assigned to delayed AI productivity.

Fiscal build-up, central case

• Public AI audits and compliance: +£0.60bn
• Regulator and enforcement capacity: +£0.25bn
• Procurement and systems changes: +£0.20bn
• Tax/receipt effects: -£0.05bn

Central net impact: +£1.0bn in 2027-28.

Behaviour and pass-through

• Low case assumes high-risk-only rules.
• Central assumes public-sector audit duties and enforcement.
• High case assumes broad human-review rights.
• Employers may delay AI deployment.
• Bias reduction benefits are not monetised.

Phasing

• 2026-27: +£0.3bn. Regulator setup.
• 2027-28: +£1.0bn. Main compliance year.
• 2028-29: +£0.9bn. Audits repeat.
• 2029-30: +£0.8bn. Systems mature.

Main source groups

• Department for Science, Innovation and Technology, "A pro-innovation approach to AI regulation" (2023): The UK AI white paper relies on principles and existing regulators rather than a single AI regulator; defines the baseline for stronger workplace AI law.
• House of Commons Library, "Artificial intelligence and employment law" (2023): Commons Library identifies employment-law issues around automated decision-making, transparency and contestability; supports worker-risk channels for AI protections.
• OECD, "Using AI in the workplace" (2024): Used to support the baseline, affected-population sizing or behavioural assumptions in the illustrative scenario.
• Acemoglu and Restrepo, "Robots and Jobs: Evidence from US Labor Markets" (Journal of Political Economy, 2020): Automation can displace tasks and workers even when it raises output in some firms; supports caution on AI rules that trade protection against productivity.
• Acemoglu, Autor, Hazell and Restrepo, "Artificial Intelligence and Jobs: Evidence from Online Vacancies" (NBER, 2022): AI exposure is visible in vacancy patterns and skill demand, not just future speculation; relevant to worker protections around AI deployment.
• Green Party of England and Wales, "Workers' Charter 2026" (2026): Used to define the pledge wording, policy scope and implementation scenario being modelled.