Fictional worked example

AI works, but the proposed reduction is too aggressive

An accounts-payable team has achieved a substantial productivity improvement from AI. Management proposes removing 30 FTE. The Stress Test supports a structural reduction, but only 18 FTE now.

You do not have to prove that AI failed to challenge the restructuring that follows.

Decision at a glance

The proposal and the evidence do not support the same number.

Proposal Remove 30 FTE now

Reduce the AP team from 100 to 70 FTE.

Capacity available for choice Approximately 18 FTE

A concentrated structural block is supported after period-end, control and coverage requirements are protected.

Supported structural outcome Remove 18 FTE now

Use redeployment, temporary assignment and attrition where they improve implementation inside the 18-position envelope.

Decision status 02: Choose a different structural option

The proposed 30-FTE immediate reduction outruns the evidence.

Employee capacity Accounts payable 100 FTE starting team AI operating at scale for 6 months

Evidence behind the case

AI productivity is real. The question is how much structural capacity it creates.

Observed evidence

  • Average invoice-processing time is 31% lower.
  • Net human workload is approximately 24% lower once exception handling, controls and remaining human work are included.
  • Approximately 18 FTE-equivalent of released capacity is concentrated in high-volume standard invoice processing.
  • Approximately 6 FTE-equivalent remains dispersed in small fractions across continuing roles.
  • Month-end and quarter-end workload has not fallen by the same amount as average workload.
  • Quality and control-error rates are unchanged.
  • The staffing model supports removing the concentrated 18-FTE block while preserving current minimum period-end coverage.

Economics and remaining assumptions

30-FTE annual saving ≈ £2.4m
18-FTE annual saving ≈ £1.45m
18-FTE one-off transition cost ≈ £0.35m
Capacity rebuild time ≈ 4 months

Assumptions still carrying the recommendation

  • Standard invoice volumes remain stable.
  • Current AI accuracy persists at scale.
  • Period-end demand does not increase materially.
  • Dispersed productivity gains may eventually consolidate, but that has not yet been demonstrated.

The capacity bridge

The gap appears before the structural decision.

The same six-stage capacity bridge used in the Stress Test shows why a 31% productivity result does not establish a 30-FTE reduction.

01

AI productivity

Average invoice-processing time is 31% lower.

02

Net human work change

Net human workload is approximately 24% lower.

03

Released capacity

Approximately 24 FTE-equivalent is released.

04

Structurally addressable capacity

Approximately 18 FTE forms an identifiable block. Around 6 FTE-equivalent remains dispersed.

05

Capacity that must remain

Period-end coverage, exceptions, controls and continuing-role requirements protect the rest.

06

Capacity available for choice

Approximately 18 FTE is available now.

30 FTE proposed 18 FTE evidenced as available for choice

The four questions

Where the proposed 30-FTE reduction survives, and where it fails.

How much human work is actually no longer required?

Supported

Average processing time is 31% lower, but that is not the number used for the structural decision. Once exceptions, controls and remaining human activity are included, net human workload is approximately 24% lower.

Human work has genuinely fallen. The proposed reduction still has more to prove.

Is released capacity structurally addressable?

Partly

Approximately 18 FTE-equivalent is concentrated in high-volume standard invoice processing and forms an identifiable block. The remaining approximately 6 FTE-equivalent is spread through continuing roles in small fractions.

18 FTE is structurally addressable now. The dispersed 6 FTE-equivalent is not six removable positions.

What capacity must remain to protect performance, coverage, skills and resilience?

18 FTE available

Period-end demand is the binding operating constraint. The retained organisation still needs capacity for exceptions, financial controls, period-end peaks, ERP-specific capability, escalation and reconciliation.

The staffing model supports removing the concentrated 18-FTE block. It does not demonstrate that another 12 FTE can leave safely.

Is the proposed reduction, at this amount and timing, economically better than the credible alternatives?

18 now is stronger

The credible choices are not only “cut 30” or “do nothing”. Different amounts, timing and implementation mechanisms need to be compared.

Option A

Remove 30 FTE immediately

≈ £2.4m annual saving

Twelve of the proposed 30 FTE are beyond the structurally available block currently demonstrated. Restoration would take approximately four months if that assumption proves wrong.

Outruns the evidence.
Option B

Remove the evidenced 18-FTE block now

≈ £1.45m annual saving · ≈ £0.35m one-off cost

The amount is supported by the staffing model and preserves required period-end coverage. Payback is roughly three months.

Strongest current structural option.
Option C

Defer the 18-FTE reduction for six months

≈ £0.12m lower one-off cost · ≈ £0.73m extra payroll

No operating requirement has been identified for carrying all 18 positions during the delay.

Delay costs more than it saves on the fictional facts.
Option D

Rely on attrition

Expected 5–7 FTE over 12 months

Timing and role mix are uncertain. Attrition can reduce implementation disruption, but it does not establish another structural tranche.

Useful implementation mechanism, not a different capacity answer.
Option E

Use some capacity elsewhere

2 permanent vacancies · ≈ 3 FTE temporary transformation work

These uses may change which people leave or the sequencing of exits, but they do not create an ongoing requirement for the full 18 AP positions.

Capture where useful inside the supported 18-position envelope.
Removing the evidenced 18-FTE structural block now is economically stronger than carrying the full block, staging the same reduction for six months, or relying on uncertain attrition to produce the right vacancies.

Stress Test conclusion

Choose a different structural option.

AI has genuinely reduced human work and a structural reduction is justified. The proposed 30-FTE immediate reduction is not the strongest available choice.

Structural decision ≠ people decision. Internal placement or temporary assignment may change which people leave. It does not change the conclusion that the AP operating model requires approximately 18 fewer permanent positions.

What could change the answer

The 18-FTE recommendation is supported now, not pre-approved forever.

Open conditions

  • Complete two quarter-end cycles after the initial reduction.
  • Track whether dispersed released capacity consolidates into identifiable structural units.
  • Monitor exceptions, control performance and period-end service.
  • Use suitable internal vacancies or temporary assignments where they improve implementation economics.
  • Recalculate minimum staffing before approving another permanent tranche.

Retest the amount if

  • Dispersed gains consolidate into identifiable capacity.
  • Sustainable vacancies emerge.
  • Period-end demand materially changes.
  • Exception or control workload changes.
  • Restoration time or cost changes materially.
  • A materially higher-value use of the capacity emerges.

What this example shows

AI can work, a structural reduction can be justified, and the proposed reduction can still be wrong.

31% faster processing 24% less human work 18 FTE available for choice 30 FTE proposed 18 FTE supported now

A productivity result does not close the gap between measured work reduction and the structural capacity the organisation can actually remove.

AI Restructuring Business Case Stress Test

Apply the same test to your own case.

Use the parent resource to test the bridge from AI productivity to a proposed permanent capacity decision.

Open the Stress Test

More worked examples

Browse all worked examples