What AI metrics should I put in front of my LPs?

Pick three or four that tie to money. Time-to-fill, recruiter output per head, gross margin per placement, and back-office cost per transaction. LPs do not want a list of tools. They want to see a number move in a direction that puts cash in the fund. Show the baseline, the current number, and the delta.

Our portcos all use different systems. Can I still measure AI across the portfolio?

Yes, but you measure outcomes, not tools. Each portco can run a different ATS and still report time-to-fill, recruiter output, and margin per placement. Normalize the definitions across the portfolio first. If one firm counts time-to-fill from req open and another counts from first submittal, your numbers are useless. Fix the definitions, then compare.

How long before AI shows real ROI at a portco?

You should see leading indicators in 60 to 90 days: faster sourcing, more submittals per recruiter, fewer manual back-office steps. Margin impact takes longer, usually two to three quarters, because it depends on recruiters actually changing how they work. If you see no leading indicators by day 90, the problem is adoption, not the technology.

PE Portfolio AI Scorecard for Staffing Portcos

Most PE-backed staffing portfolios can't prove their AI spend is paying back. Here is a portfolio-wide scorecard you can build before your next LP report, using numbers you already have.

Most PE-backed staffing portfolios are spending on AI and cannot prove it is working. That is the quiet problem behind the next LP report. The deck says the portfolio is adopting AI. The numbers behind it do not exist, or they live in five different formats across five portcos that count things five different ways.

The scale of the gap is real. StaffingHub reported 35 staffing transactions in Q1 2026, the strongest opening quarter in at least three years, with IT staffing and executive search making up nearly half of all activity. Every one of those deals creates a new portco with its own systems, its own data, and its own idea of what AI is supposed to do. If you run ops across that portfolio, you are now responsible for proving AI pays back across companies that do not even share a common dictionary.

The fix is not another tool. It is a scorecard. One page, built from numbers you already have, that lets you compare AI impact across every portco and walk into the LP meeting with proof instead of vibes.

Why the Current Reporting Falls Apart

Ask ten portco leaders how AI is going and you will get ten stories. One is excited about a sourcing tool. One bought something and never rolled it out. One has recruiters quietly using AI on their own phones. None of that is measurement. It is anecdote dressed up for a board deck.

The reason this happens is simple. Most portfolios skipped the boring step: defining what a good outcome looks like before they bought anything. So when the LP asks "what is the return," the honest answer is nobody knows. The spend is visible. The payback is not.

A scorecard forces the discipline the buying process skipped. It makes each portco report the same three or four numbers, measured the same way, on the same cadence. That is the whole trick. Not fancy dashboards. Shared definitions and a short list of metrics that tie to cash.

The Four Metrics That Actually Belong on the Scorecard

You do not need twenty metrics. You need four that a limited partner understands without a translator. Here is the set I put in front of ops leaders.

1. Recruiter output per head

How many submittals, interviews, or placements does one recruiter produce in a month? This is the clearest sign AI is doing real work on the desk. If a recruiter was placing four a month and now places six because AI handles sourcing and screening grunt work, that is a number an LP feels in their gut. Track it per portco and per team so you can see where adoption is real and where it is theater.

2. Time-to-fill

From req open to candidate accepted. AI that works compresses this. AI that sits unused does not. Set one definition for the whole portfolio. If one firm starts the clock at req open and another starts it at first submittal, you cannot compare them, and your portfolio number is fiction.

3. Gross margin per placement

This is the one LPs care about most, so it goes on the page. AI should either raise the bill rate through faster, better matches or lower the cost to fill through less manual work. Either way, margin per placement moves. If it is flat after two quarters of AI spend, you have an adoption problem, not a technology problem.

4. Back-office cost per transaction

Onboarding, timesheets, payroll, invoicing. AI in the back office shows up as fewer hours spent per transaction and fewer errors. Pick one or two high-volume processes per portco and measure the cost to run one unit. When that number drops, you have hard savings you can defend.

Four metrics. Each one ties to money. Each one works across portcos that run different systems, because you are measuring the outcome, not the software.

How to Build the Scorecard in Two Weeks

This does not require a consulting army or a six-month project. It requires a week of definition work and a week of data pulling.

Week one, lock the definitions. Get one leader from each portco on a call. Agree on exactly how each of the four metrics is counted. Write it down. Send it back. This is the step everyone wants to skip and the step that makes the whole thing valid.
Week one, set the baseline. For each portco, pull the four numbers for the last full quarter before AI went in. If AI is already in use, pull the quarter before the rollout started. No baseline means no story.
Week two, pull current numbers. Same four metrics, most recent quarter. Put baseline and current side by side. The gap is your ROI story.
Week two, flag the adoption gaps. Where a number did not move, do not hide it. Ask why. Usually it is adoption. Recruiters bought the tool and kept working the old way. That is a people problem, and it is fixable once you name it.

That last point matters more than any metric. Technology does not produce ROI. People using the technology differently produces ROI. If your scorecard shows flat numbers at a portco that spent real money, the issue is almost never the product. It is that nobody changed how the work gets done. That is where real change management earns its keep.

What the Scorecard Tells You That a Tool List Never Will

Once you have one page with four metrics across every portco, patterns jump out. You will see the portco where recruiter output jumped 30% and you can study what they did and copy it. You will see the one that bought the most and moved the least, and you can go fix adoption instead of buying more. You can point to the portfolio-wide delta in margin per placement and say, this is what the AI spend returned.

That is a different LP conversation. Instead of "we are investing in AI," you get to say "AI added this many basis points of margin across the portfolio, here is where, and here is the one portco we are turning around next quarter." Specific beats aspirational every time in that room.

The portfolios that win the next few years will not be the ones with the most tools. They will be the ones that can prove, portco by portco, that the tools pay back. The scorecard is how you prove it.

Do this week: Pick one portco. Pull the four metrics for the last full quarter before AI went in, and for the most recent quarter. Put them side by side on one page. If the numbers moved, you have the start of your LP story. If they did not, you just found your adoption problem before the LP did.

The PE Portfolio AI Scorecard: How to Measure AI ROI Across Staffing Portcos Before Your Next LP Report