Cost out, done properly, means removing the work that generates cost, not removing the people who do it. Most PE cost-out programs do the second thing, book the savings in the first hundred days, and watch the cost quietly reappear before the next audit. The line item comes back because the work never left. It just got redistributed to whoever was still standing, and then a backfill req shows up two quarters later with a straight face.

If you are the operating partner, this is your call and your credibility. You have a value-creation plan with a cost number on it, a management team that will nod in the meeting and slow-walk it after, and an IC that will ask at exit why the margin gains didn't stick. This page is about the difference between a cost cut that survives to exit and one that shows up as an add-back nobody believes.

Why does the cost come back?

Because a headcount cut is a bet that the work will get done with fewer hands, and that bet is usually wrong in an operations-heavy business. The work is a chain. Quote to cash, order to invoice, claim to payment, spec to build. Each chain is thirty to fifty manual steps, and most of those steps are the same shape: a person reads something, checks it against something else, types it into a third system, and moves it along. Cut the people and the chain doesn't shorten. It just runs slower, breaks more, and the breaks land as write-offs, DSO creep, and rework that never appears on the cost-out scorecard because it hides in a dozen other cost centers.

Here is the operating reality specific to a portco: the most junior, cheapest people in the building run the most expensive processes. One wrong checkbox in a quote, a mis-keyed rebate, a missed contractual term, and you have a six-figure leak. When you thin the ranks under a cost-out mandate, you don't remove that risk. You concentrate it. Fewer people, same volume, more pressure, more errors. The savings on the payroll line get eaten by the error rate you can't see.

So the honest diagnostic is not "where can we cut heads." It is "where does a person's hands on a keyboard produce nothing but risk." Those steps are pure cost with no judgment in them. That is where cost comes out and stays out.

Where does the cost actually hide?

Walk one chain end to end before you touch a spreadsheet. Not the org chart, the chain. Pick your most manual, highest-volume process and count the discrete steps a human touches between the trigger event and the finished output. In the businesses we have rebuilt, that count is routinely thirty to fifty. Then sort each step into one of two buckets: judgment, or transcription.

Judgment is deciding what a customer should be quoted, whether a claim is valid, how to sequence a build. That is where your experienced people earn their salary and where the cost is worth paying. Transcription is everything else: copying a value from one screen to another, reconciling two lists, formatting a document, chasing a status. In a real operation, transcription is the majority of the steps and a large share of the fully loaded cost, and none of it needs a human.

The number that matters is not FTE count. It is the volume of transcription events per year. On a PE-backed demand-generation business, a fifty-step quote-to-cash process collapsed onto an agent-run spine and removed 12,450 manual sourcing events a year. That is cost out that holds, because those events are simply gone. Nobody backfills a step that no longer exists.

The forecasts on how much AI "could" save an industry disagree with each other and the dollar figures are noise. Ignore them. Count your own transcription volume. That number is real and it is yours.

How do you take the cost out without breaking the operation?

Sequence it so the COO stops holding their breath. The failure mode every operator has lived is the new ops lead who maps the whole operation, restructures it, finds the real problem, and leaves before the fix ships. The map was right. Nobody carried it to production. Cost out fails at the same handoff.

The sequence that works:

First, instrument one chain, not all of them. Prove the transcription steps can move to agents while the experts keep the judgment calls. On a PE-owned automotive-software platform, integration classification ran at ninety percent accuracy and data coverage moved from 53 percent to 81 percent, which is what "it actually runs" looks like in numbers, not in a demo.

Second, run every change in a live environment on a real copy of the data and replay every check before anything ships. This is the discipline that separates a cost-out program from a science project. If a change can't pass the replay, it doesn't go live, and the operation never gets disrupted to find out.

Third, hand the running system to the portco and keep the experts on the judgment. The people you kept now spend their day on the decisions that actually move margin, not on keying data. That is where the second-order EBITDA comes from: not just the removed cost, but the recovered capacity.

Do not start with the tool. The model is about five percent of the gap. The other ninety-five is who owns the judgment when the chain is running and something unusual arrives. Name that owner before you write a line of the plan.

Is this predictable, and who carries the risk?

The CFO's question is the right one. A cost-out program that overruns its own budget is worse than no program, and the history of consulting engagements and stalled AI pilots gives every reason to be skeptical. Predictability comes from two things: a fixed scope tied to one chain with a countable outcome, and the replay discipline above that makes a fixed price possible in the first place. You can price an outcome when every change is validated against real data before it ships. You cannot price "transformation."

The risk you should refuse to carry is the one where you pay for effort and hope for a result. Tie the money to the KPI, put a date on it set with the scope, and make sure someone senior owns the number end to end. If the removed cost doesn't materialize, that is the vendor's problem to keep working on, not a line you already expensed.

How Salfati Group would approach this

We take cost out as a Mandate: fixed price, fixed scope, anchored to a named KPI such as manual events removed or cost per transaction, backed by an Outcome SLA. If we miss the target, we keep working at no additional cost until the number ships. A named architect owns it end to end, agents do more than eighty percent of the work, and the portco owns the system that ships, not a dependency on us. The cost comes out of the chain, which is why it holds through exit instead of reappearing as a backfill. If you have a chain in mind and a number on it, start at /apply.

Sources

1. 1. [PDF] Uncovering the Costs and Benefits of Private Equity - StepStone Group» Organizational expenses generally include the out-of- pocket expenses incurred when forming the fund and any related vehicles, such as printing, travel ... Private Equity fees are abundant in that the headline fee levels are high, especially relative to public index strategies. ... private equity...
2. 2. [PDF] In this research paper, we express private equity fees in Total ...In total, an investor with a $100 portfolio pays $2.8 p.a. in management fees, implying an MER of 2.8% pa. Based on the return assumptions set out in Appendix 1 ... between 1.25%-2.0% of committed capital for the first 5 years and then net invested capital thereafter but there are several variants ...
3. 3. What you should know:Evaluating Private Equity Fees - Hamilton Lane There are three main fee types that PE firms and GPs use to align their success with the success of the portfolio companies in their funds. Below are three types of fees that may impact returns. | |**Calculation**|**% of Committed Capital, Net Asset Val...
4. 4. The Impact of the Rising Cost of Debt on Private Equity - CommonfundAs the cost of debt on private equity (PE) has risen and its availability has fallen, deal-making within PE has slowed. ## As the cost of debt on private equity ("PE") has risen and its availability has fallen, deal-making within PE has slowed. ... This routinely yielded positive outcomes when debt ...
5. 5. What is the Cost of Private Equity? - Zachary ScottGrossing up this return for management costs implies a cost of equity in the range of 18% for a private equity investment. After adjusting for these factors, our calculations lead us to believe that the required return-on-equity for similar businesses should be much lower. A 15%-20% return on equity...
6. 6. [PDF] The Economics of Private Equity FundsThe most common initial fee level is 2 percent, though the majority of funds give some concessions to LPs after the investment period is over; e.g., switching ...
7. 7. Private Equity: What You Need to Know - KKRUnderstanding Private Equity Fund Fees. PE firms charge investors two types of fees: a management fee, which is typically between 1% and 2% of committed capital ... PE has, over the long term, outperformed public markets. Over the past 25 years, PE has delivered 5% or ~500 bps on average more than g...
8. 8. The Dark Side of Private EquityThis article contends that the PE investment model imposes social costs in many portfolio companies through over-leverage, value extraction and short-term ... Private equity (PE) funds control over $9 trillion in assets and thousands of companies, yet their leverage-driven model often amplifies fina...

Reviewed by David Fialho·25 August 2026