Fictional worked example

AI works, but no capacity is released

A 240-FTE customer-operations function has reduced human handling effort per transaction by approximately 35%. But transaction volume is 45% higher and the AI-enabled model creates new human work. Management proposes removing 36 positions. The Stress Test supports zero permanent reduction now.

AI productivity can be real while net human workload does not fall at all.

Decision at a glance

The unit-productivity gain has been absorbed by higher demand and new AI-related work.

Proposal Remove 36 positions

Management treats a 15% staffing reduction as conservative against a 35% productivity improvement.

Released capacity Approximately 0 FTE-equivalent

Current human workload is approximately unchanged after demand growth and new AI-related work are included.

Supported structural outcome 0 permanent reduction now

Continue approximately the current staffing structure while capturing the AI benefit through higher throughput and avoided growth hiring.

Decision status 02: Choose a different structural option

The proposed 36-position reduction has no current capacity basis.

Employee capacity Customer operations 240 FTE starting team AI operating at scale for 9 months

Evidence behind the case

Each transaction needs less human effort, but the organisation now has far more transactions to serve.

Observed evidence

  • Before AI, the workload required approximately 240 FTE-equivalent at the old transaction volume.
  • Human handling effort per transaction is approximately 35% lower.
  • Transaction volume is approximately 45% higher.
  • The new operating model requires approximately 14 FTE-equivalent of exception review, AI quality sampling, escalation, controls, knowledge-base maintenance, monitoring and governance.
  • Total current human workload is approximately 240.2 FTE-equivalent.
  • Service levels remain inside target.
  • Quality is stable.
  • The organisation has avoided substantial growth hiring.

Workload arithmetic

Old workload ≈ 240 FTE-equivalent
35% lower effort at old volume ≈ 156 FTE-equivalent
After 45% demand growth ≈ 226.2 FTE-equivalent
Plus new AI-related work ≈ 240.2 FTE-equivalent

What the AI has achieved

  • Approximately 45% more transactions with essentially the same human capacity.
  • Lower unit operating cost.
  • Stable service at higher volume.
  • Avoided growth hiring.
  • Economic value without a released structural capacity block.

The capacity bridge at the stopping point

The productivity gain does not become released capacity.

A 35% reduction in human effort per transaction does not become a capacity percentage when total human workload remains approximately flat.

01

AI productivity

Human effort per transaction is approximately 35% lower.

02

Net human work change

After 45% demand growth and new AI-related work are included, total human workload is essentially unchanged.

03

Released capacity

Approximately 0 FTE-equivalent is durably released at the current demand level.

Stop after Q1 No relevant released capacity to carry into Q2

The stopping point

The analysis stops after Q1.

Q1 establishes approximately zero released capacity. There is no relevant capacity to carry into Q2, so Q2, Q3 and Q4 are not responsibly reached.

How much human work is actually no longer required?

Approximately none

The 35% per-transaction improvement is real. But after 45% demand growth and approximately 14 FTE-equivalent of new AI-related work are included, required human workload remains approximately flat.

AI has improved productivity without producing a material net reduction in human workload. That translates to approximately 0 FTE-equivalent of released capacity at the current demand level.
Stop Q2, Q3 and Q4 are not responsibly reached on the current evidence.

Stress Test conclusion

Choose a different structural option.

The proposed 36-position reduction is not supported. The workload evidence supports approximately zero permanent reduction now.

Zero permanent reduction is a structural decision. Do not pre-authorise a future reduction based on capacity that does not exist today.

What could change the answer

Retest only when new facts change the net workload or structural capacity.

Retest if

  • Demand falls materially.
  • Handling effort declines further.
  • Exception or control work decreases.
  • Another workflow redesign consolidates genuinely surplus work.
  • Service requirements change.

Until then

  • Operate at approximately the current staffing level.
  • Capture the AI benefit through higher throughput.
  • Recognise avoided growth hiring as real economic value.
  • Maintain service and quality at the higher transaction volume.
  • Do not convert the 35% unit-productivity improvement into a headcount percentage.

What this example shows

AI can succeed without releasing any structural capacity at all.

AI productivity improves materially demand rises new AI-related work is added net human workload remains approximately flat released capacity is approximately 0 STOP 0 positions now is the evidence-supported structural result 02: Choose a different structural option

The AI has created real economic value. It has not created removable capacity. A productivity percentage is not a capacity percentage.

Free Execution Pack

Apply the Stress Test to your case.

Download the Word analysis, one-page PowerPoint executive summary and User Guide to analyse, challenge and document a live AI-backed structural capacity decision.

Explore the Execution Pack
Learn about the AI Restructuring Stress Test Browse more examples