AI can do the number-crunching here, but a consultant has to own the diagnosis.

The analytical framework (COGS/SG&A/asset utilization decomposition, ratio benchmarking) is structurally repeatable across engagements. However, each client's data quirks, industry sub-segment, and competitive context require bespoke interpretation, reducing true repeatability.

Success criteria are deceptively vague: 'highest-impact levers' and 'quantified EBITDA upside' sound crisp but depend on assumptions about implementation feasibility, peer benchmarks, and client appetite that no agent can resolve from the data alone. The agent cannot know when it's done well enough.

The client has provided structured financial and headcount data, which an agent can ingest and analyze. However, meaningful benchmarking requires external industry comps (automotive tier-2 EBITDA margins, labor cost norms) that the agent may not have reliable access to without curated databases.

A miscalibrated EBITDA upside estimate directly shapes the engagement scope and fee proposal — an error could cause the firm to underprice, overpromise, or lose credibility with a $180M client. Errors here are reputationally and commercially costly and not easily reversed once the proposal is submitted.

Identifying which levers are genuinely actionable versus theoretically attractive requires operational intuition about automotive supply chains, labor relations, and capital constraints. The consultant also needs to calibrate findings to what the client will actually accept — a judgment call no agent can make from financials alone.