How does AI move EBITDA in a lower-middle-market portco?

For a management team working within a tight budget, the most useful AI discussion begins with an operating problem: a growing service backlog, slow quote turnaround, expensive rework, or administrative demands that keep experienced employees away from customers.

Each problem has a cost. Each also has constraints that technology alone may not resolve.

The sponsor's task is to help management select opportunities where a measurable operating improvement can produce a credible financial return.

A useful AI business case identifies the financial line it is intended to change.

If the objective is lower customer service expense, establish the current cost of handling the workload and the portion of that cost management can realistically reduce.

If the objective is additional sales, establish where business is being lost and whether the company can fulfill more orders profitably.

If the objective is fewer errors, identify the credits, write-offs, rework, or other costs those errors create.

This discipline helps a team choose among competing opportunities. It also prevents an appealing demonstration from becoming a project without a clear economic purpose.

AI may help employees retrieve information, prepare documents, summarize interactions, or process incoming requests more quickly.

Research supports the potential for productivity improvements in specific settings. In a study of customer support agents, researchers found that access to an AI assistant improved productivity, with benefits varying across workers. That finding supports testing a relevant workflow; it does not establish the likely return for a particular portfolio company. See "Generative AI at Work," published through NBER.

For management, the next question is what happens to the capacity created.

A service team may absorb higher volume without adding staff. A finance department may reduce outside processing costs. An operations group may bring down overtime.

Each outcome requires an explicit decision. If staffing, external spending, and profitable output all remain unchanged, the productivity improvement has not yet produced an EBITDA benefit.

Avoided hiring also needs careful treatment. It can improve performance against a credible growth budget, but it is different from reducing expenses already present in historical earnings.

A company may have more demand than its existing processes can accommodate.

Quotes take too long to prepare. Qualified inquiries sit unanswered. Experienced staff spend hours assembling information before they can advise a customer.

AI may help shorten those steps, allowing the business to handle more opportunities with its existing team. The financial result still depends on conversion, pricing, fulfillment capacity, and the incremental cost of delivery.

Suppose faster quoting leads to additional orders. The appropriate EBITDA contribution is the profit remaining after the incremental costs of winning and serving those orders. Counting the full revenue increase would overstate the benefit.

Management should also verify that the activity is incremental. Orders moved forward from next month do not necessarily represent additional annual revenue.

Some opportunities sit in the gap between work performed and value collected.

Examples include missed billable services, inconsistent application of contract terms, avoidable service credits, and incomplete documentation that delays resolution of a customer dispute.

AI can help identify exceptions or prepare information for review. Employees still need the authority and time to resolve them.

Be precise about the outcome. Faster collection of an existing receivable improves working capital. It does not, by itself, increase EBITDA. Recovering valid, previously unrecorded revenue or preventing an avoidable expense may affect earnings, subject to the company's accounting treatment.

Finance should establish that distinction before benefits enter the reporting package.

Consider an illustrative portfolio company with $40 million in revenue and $5 million in EBITDA. These figures are hypothetical, not client results or a forecast.

Assume management validates the following annual benefits once the new workflows are fully operating:

On these assumptions, annual run-rate EBITDA increases by 7%, from $5 million to $5.35 million.

The calculation assumes the benefits are distinct, achievable, and measured against a consistent baseline. Counting the same employee capacity as both labor savings and support for additional orders would overstate the result.

Timing matters, too. If the company realizes only half of the net annual benefit during the first year and incurs $120,000 in implementation expenses charged against EBITDA, the first-year improvement is $55,000.

Management should show first-year results, steady-state economics, and cash investment separately. A compelling annual run rate can still require patience and funding during implementation.

For a lower-middle-market company, a focused starting point is one workflow with meaningful volume, accessible data, and a business owner who can change how the work gets done.

Before launch, agree on the baseline, quality requirements, full costs, and criteria for expanding or stopping the initiative.

During the pilot, measure adoption and exception handling alongside speed. Include the time employees spend checking and correcting outputs. A faster first step offers limited value if it creates more work downstream.

At the end of the period, finance and the operating owner should be able to explain what changed, how much value was realized, and what remains unproven.

The board should receive a concise account of the operating result, its financial effect, and the decision management recommends next.