Most private equity firms now have an AI position. Fewer have an AI result they can point to in a portfolio company's numbers. The gap is not conviction or budget. It is sequencing: which company, which function, which use case first, and how the second and third inherit from the first.
Start with the map
A use-case map is a short exercise across the portfolio: for each company, the two or three places where a working system would move a P&L line the operating partner already watches, sized in the company's own numbers. It takes a day or two per company and produces a ranked list across the fund, with the payback, the systems involved and the risks stated plainly. It replaces twenty uncoordinated pilots with a portfolio-level decision.
The patterns that repeat across a portfolio
Because portfolio companies share the same operating shapes, the map usually converges on a handful of use cases: technical sales and order intake in distributors, invoice audit against contracts in anything with high supplier counts, long-tail procurement negotiation in manufacturers, claims and service triage in logistics, and variance narratives in every finance function. A system built well once can be customised for the next company in a fraction of the time, because the objects and rules are largely the same.
Sequence for compounding, not coverage
The temptation is to run a pilot in every company at once. The better pattern is to go deep in one or two, prove the number, and build the model that the rest of the portfolio can reuse. A second company inheriting a proven use case and a proven operating system ships in weeks and carries far less execution risk. Coverage comes faster by compounding than by parallel starts.
Measure the way an investment committee measures
Every system gets a baseline before it is built, a target the management team signs, and a performance gate at month six. Results are reported on the P&L line they were built for, not as "AI initiatives". Over a hold period that produces something rare: a value-creation story a buyer can verify in the numbers, and an owned asset, the operating system, that transfers with the company at exit.
What the fund owns at the end
A repeatable playbook: a ranked map, a set of proven systems, a delivery method with gates, and a reference point in the numbers for the next deal. The portfolio companies own their models and their data. The fund owns the pattern. Both compound.