An AI demonstration can look impressive and still produce very little business value. A task that once took an hour now takes ten minutes. Everybody likes the result. Then somebody asks what changed in the financials, and the room gets quiet.
I would start an AI investment discussion with that last question. For a private equity portfolio company, the useful measure is what the business can do with the capacity it gets back. That could mean handling more customers with the same team, improving response times, avoiding a planned hire, or reducing an actual expense. Each is a different business case.
Separate capacity from cash
Saved employee time is capacity. It becomes a financial benefit when management makes a decision about how to use it.
Take a hypothetical customer operations team processing 1,000 requests a month. Suppose AI reduces handling time by 12 minutes per request. That is 200 hours of potential capacity. At an assumed loaded cost of $40 an hour, the spreadsheet shows $8,000 a month.
That calculation does not establish an $8,000 expense reduction. If payroll stays the same, the team has gained capacity. If reviewing the AI output adds five minutes per request, the net time benefit is much smaller. If the team uses the capacity to support growth, the economic result depends on whether demand exists and the additional work produces contribution margin.
I want all three numbers visible: gross time saved, net capacity released, and the benefit actually realized. Mixing them makes a pilot look better than the business.
Put the full cost in the decision
The subscription is usually the easiest cost to see. Implementation, integration, data preparation, employee training, output review, and ongoing support belong in the same calculation.
For the hypothetical team, assume the expected time saving is confirmed after review. If monthly software, support, and review costs total $3,000, the project still needs a plan for using the remaining capacity. An avoided hire can be a credible benefit if that hire was genuinely planned and the workload can be covered. It should be identified as cost avoidance, with its timing and assumptions recorded.
Revenue benefits need equal discipline. Faster quotes may help win business, but quote speed alone does not prove incremental sales. Track conversion, gross margin, and whether faster turnaround changes customer behavior. Avoid assigning the same revenue improvement to several initiatives.
Give the benefit an owner
Technology teams can own deployment. The executive who controls the workflow should own the operating benefit.
Before approving a pilot, I would ask for a short benefit plan:
- The workflow and its current volume, cost, quality, and turnaround time.
- The change being tested, including human review and exception handling.
- The person responsible for putting released capacity to work.
- The evidence needed to continue, expand, revise, or stop.
A finance partner should agree with the measurement method before the project starts. That avoids a successful pilot being followed by an argument about what success meant.
Review the business result after deployment
Measure the workflow under normal operating conditions. A few clean examples selected for a demonstration will not tell you what happens with incomplete records, unusual requests, or a busy Monday morning.
My review would include adoption, net handling time, rework, customer outcomes, and the disposition of saved capacity. If employees spend the recovered time checking a second system or repairing mistakes, the original estimate needs to change. If throughput improves without additional headcount, document the demand served and the incremental margin rather than treating every recovered hour as cash.
For sponsors managing several portfolio companies, a common measurement format is useful. The use cases can differ while the distinction between capacity, cost avoidance, actual savings, and revenue contribution stays consistent.
AI can make a good team more productive. Management still has to decide how that productivity improves the business. I would make that decision part of the investment case from day one.
Related reading: A 90-day AI roadmap and AI agents and operating controls.
If the AI business case needs an operating owner, see how I work with leadership teams.