Practice interviews

Would you underwrite the portfolio AI plan?

Distinguish tested savings from a business model that customers may stop paying for.

“growth underwrites are central to anything that comes through our investment committee.” (opens the source)

Dipan Patel, Co-Managing Partner and Co-CEO, Permira · Goldman Sachs Exchanges: Great Investors (published transcript), Sep 4, 2026

The question

You are reviewing fictional Meridian Services, a UK compliance-research business. Management proposes an AI research assistant as part of a private equity investment case.

Management proposal
USD millions
Proposed annual staff savings4
Annual software and review cost1
Upfront rollout cost2
Routine-summary revenue exposed12

Invented proposal, not proven savings or an assumption that all exposed revenue is lost.

The evidence
Demand risk
Clients in the exposed routine-research segment have not agreed to renew if the service changes.
Evidence
A small internal trial cut drafting time. It did not test client acceptance or the cost of specialist review.

“What goes into the base case, what stays upside, and what would make you pass?”

A strong answer, step by step

Open each step in turn. This is one strong answer, not the only one.

Step 1: Rebuild the cost bridge

What would you do here? Open the move you’d make.

  • Put USD4m of savings into the base case.

    Tempting, but it misses something. That omits recurring cost, implementation cost and customer risk.

  • Separate proven operating results from proposed savings.

    The strong move. The proposal subtracts recurring cost but has not tested the full operating model.

See how a strong candidate says it

“The claimed steady-state net saving is USD3m before demand effects, with USD2m upfront. It is not yet a proven base-case saving.”

What a strong candidate could say

Why it works: The proposal subtracts recurring cost but has not tested the full operating model.

Step 2: Test the revenue at risk

“Segment the USD12m of routine-summary revenue. Interview renewing clients and test what they will still pay for after AI changes delivery.”

What a strong candidate could say

Why it works: A cost reduction can destroy value if it weakens the paid customer need.

Step 3: Stage the investment

“Fund a bounded client pilot with specialist review. Require evidence on error rates, renewal intent and net delivery cost before scaling.”

What a strong candidate could say

Why it works: Internal drafting speed does not prove commercial adoption.

Step 4: Name the pass condition

“Pass if the investment only meets its return target with unproven savings or a stable revenue line contradicted by client evidence.”

What a strong candidate could say

Why it works: An investment recommendation must say which assumptions it cannot carry.

The answer, in short

Decision
Keep untested savings outside the base case. Underwrite client demand and review cost before the full rollout.
Because
The internal trial says little about the USD12m revenue exposure or the specialist review burden.
What would change it
Client renewals and independently measured net delivery savings could support a staged base-case benefit.

Common traps

  • Adding gross staff savings without review cost.
  • Calling all exposed revenue lost.
  • Assuming faster research is a product buyers still value.

Check your answer

  • Did I calculate net recurring savings?
  • Did I treat upfront cost separately?
  • Did I test demand as well as efficiency?
  • Did I state a reason to pass?

Last reviewed Nov 10, 2026