Human by choice · Worked analysis · A Policy Window advocacy programme
Who pays to keep the human time?
Prepared Published Edition 2
An AI tool might leave staff more time for people while adding a bill the service cannot fund. A supplier's price, the buyer's VAT recovery and usable tax relief are different things. Before promising that technology will pay for better human contact, ask whose budget actually gains money.
This extends our invented example in When AI saves time, who gets it?. It is an illustration, not evidence of AI performance or advice for a real investment or tax return.
Keep the service in view
The service provides 1,000 appointments per period. Manual work takes 30 minutes each: 20 in conversation and ten in administration. Assisted work takes 25: the same conversation plus five minutes for checking and administration. Five net minutes saved after necessary checking is an assumption. Equal service output is a target; equivalent quality, safety and access have not been demonstrated.
Manual work totals 500 hours; assisted tasks take 416⅔, potentially releasing 83⅓. The original £40 assigned hourly time value gives that capacity a £3,333⅓ value. It is not a newly specified gross wage or proof that any payroll can disappear.
Here we add an explicit price condition absent from the original: domestic, standard-rated software and setup charges of £4,000 before VAT in a setup period, or £2,000 in a recurring period. These are invented prices; unlisted real costs still need investigation.
HMRC gives a 20% standard VAT rate. Recovery depends on eligibility and evidence, including the buyer's activities; registration alone does not settle every purchase. [1][2]
The same invoice, different budgets
For this table, each fictional buyer employs and pays the staff and buys the technology. A recovers all VAT and has common, immediately usable 25% profit-tax relief on both qualifying costs. B has neither in this period. Neither describes a sector. [2][3]
A's setup deduction is stipulated. Actual software classification and timing need checking; B's possible later relief is unmodelled. [4]
Both pay a gross £4,800 setup-period invoice or £2,400 recurring invoice. A recovers £800 or £400 VAT; B recovers nothing. A recovery or tax reduction received later cannot pay today's invoice without interim funding.
Now compare retaining paid costs with stipulating that all £3,333⅓ becomes avoided expenditure. The latter is demanding, not a finding that reduced paid hours are feasible, lawful or desirable. Overheads may remain; workers might lose income. A real plan must establish the authority for any pay or staffing change. We do not assign any saved minutes twice.
| Buyer and paid-cost choice | Setup: before profit-tax effects | Setup: after stipulated relief | Recurring: after stipulated relief |
|---|---|---|---|
| A: retain costs | £4,000 cost | £3,000 cost | £1,500 cost |
| B: retain costs | £4,800 cost | £4,800 cost | £2,400 cost |
| A: full avoidance | £667 cost | £500 cost | £1,000 saving |
| B: full avoidance | £1,467 cost | £1,467 cost | £933 saving |
Increment against manual operation, after VAT recovery. Rounded pounds; exact-fraction calculation. Each period uses fixed conditions, without forecasting law.
For example, A's recurring £2,000 bill minus £3,333⅓ avoided spending leaves a £1,333⅓ saving before profit-tax effects. The net saving is £1,000 because avoiding deductible expenditure also removes its tax deduction. Relief does not turn retained salaries into cash savings.
Follow the funding to the person
Retaining paid staff time in this example needs additional funding. The employer, commissioner buying the service and recipient of a grant may control different budgets. A benefit in one account cannot finance another without an actual transfer or funding agreement.
Use the existing business case to identify the employer and purchaser, the invoice payer, who receives recovery or relief, and who can move the resulting funds. Record invoice and receipt dates alongside the amount available for staff. Workers and service users should help decide what retained human work is for; a budget calculation supplies no mandate to cut it.
Compare the proposed tool with the strongest feasible alternative: a better manual process, different staffing or training. Include transition and evaluation costs. Review the decision if net time gains fail, errors or access worsen, or the funding assumed for human work does not arrive. The same service target needs testing under every option.
Reindustrialise UK's broader analysis examines taxation and automation across the economy. Our narrower account cannot establish who ultimately bears a tax, whether automation improves social welfare, or which tax policy is best. Productive technology may complement people and justify extra spending. A lower buyer's bill alone settles none of those questions.
Keeping human work is partly a question of power over money. An honest promise names who can pay, when the funds arrive, and who decides how they are used.
Human by choice is Policy Window's advocacy programme on AI, work and who decides. This analysis and advocacy is distinct from the neutral Policy Window wiki.
Sources
Current guidance checked 12 November 2026; the worked service and buyer conditions are fictional.
[1] HMRC: VAT rates.
[2] HMRC: reclaiming VAT.