On September 16, the Federal Open Market Committee voted 12 to 0 to raise the federal funds target range by a quarter point, to 3.75% to 4.00%. The statement was short and unambiguous: “Inflation remains elevated.” The Committee said the move “will support a timelier return” to its 2% goal.¹
For private markets, a higher policy rate shows up in familiar places. Financing costs go up, the hurdle for every new allocation rises, and LPs look harder at what they are paying for. It also shows up somewhere managers talk about less: the management company’s own budget.

When money was cheap, it was easy to add tools on a hunch. At these rates, every recurring expense in a small GP’s operating budget should be able to explain itself. AI included.
“AI-Powered” Is Not a Strategy, and Examiners Have Noticed
There is a second reason to be disciplined. In its Fiscal Year 2026 examination priorities, the SEC’s Division of Examinations said it will “review for accuracy registrant representations regarding their AI capabilities,” and assess whether firms have adequate policies and procedures to monitor and supervise their use of AI, including in back-office operations.²
Read that alongside your pitch deck. If the deck says your process is “AI-driven,” you should be able to show, specifically, what the AI does, who supervises it, and how you know it works. A vague claim that you can’t back up is now a compliance question as well as a marketing one.
Our view is that the right amount of AI in a manager’s marketing is roughly zero, and the right amount in operations is whatever you can measure.
Where the Operating Leverage Actually Is
The work in a small fund that AI handles well tends to share three features: it is repetitive, it is text-heavy, and its correctness can be checked. In our experience, the best returns come from a short list of workflows.
DDQ and RFP responses. Many LP questions repeat with small variations. A source-linked answer library, maintained by AI and approved by a person, can turn a multi-day scramble into a review session.
LP and pipeline follow-up. Turning meeting notes and emails into dated, owned next steps is tedious for people and straightforward for a well-designed system.
Data hygiene. Duplicate contacts, stale CRM stages and mismatched figures across documents are expensive in ways that rarely show up on a budget line until an LP points one out.
What is notably absent from the list: investment decisions, valuation judgments and anything that relies on a relationship. AI can prepare the materials. It shouldn’t make the call.
A Lesson From Our Own Bill
We run our firm on an internal AI operating system, and the most useful cost lesson we learned had nothing to do with which model we used.
One of our periodic deep runs once took close to three times its normal duration. The cause was not a hard problem. The system was re-reading material that hadn’t changed since the previous run: whole message threads, full transcripts, entire lists. The fix was to read only what changed since the last checkpoint, reuse summaries of material it had already processed, and give each stage a time budget.
The broader principle is to design for the smallest unit of work. A system that re-reads everything to find one change is slow, and in AI, slow usually means expensive. Processing only the items that changed is usually both faster and cheaper, which is not what most people expect when they hear “more automation.”
How to Budget AI Like a CFO Would
If you are deciding where to spend on AI this quarter, here is the approach we recommend to clients:
- Measure the baseline first. Hours per week on each candidate workflow, who does it, and what errors cost when they happen. Without a baseline, any result looks like success.
- Price it per task, not per seat. What does one DDQ response, one LP follow-up or one data-room refresh cost before and after? Per-task numbers make trade-offs obvious.
- Run one workflow for 60 days. Then keep it, fix it or kill it. Pilots that never end are a quiet drain.
- Count the review time. If a person has to check every output line by line, the savings are smaller than they look. Good systems make review fast by linking every output to its source.
- Write down what the AI does. One page, plain language, reviewed by whoever owns compliance. It doubles as your answer when an LP or an examiner asks.
The Bottom Line
Higher rates don’t make AI less useful. They make waste more visible. The managers who come out ahead will be the ones who can point to specific workflows, specific hours saved and specific controls, and who say less about AI in their marketing because they have more to show in their operations.
That is the lens we bring to Business Process AI (BPAI) engagements: start with the process, measure it, and only then decide where AI belongs.
Find Where AI Pays for Itself in Your Firm
We can help you baseline your operating workflows, pick the one or two where AI earns a measurable return, and document it in a way LPs and examiners can follow.
References
- Board of Governors of the Federal Reserve System. “Federal Reserve issues FOMC statement.” September 16, 2026. https://www.federalreserve.gov/newsevents/pressreleases/monetary20260916a.htm
- U.S. Securities and Exchange Commission, Division of Examinations. Fiscal Year 2026 Examination Priorities, section on Emerging Financial Technology. https://www.sec.gov/files/2026-exam-priorities.pdf
Start with the workflow, not the tool. See our guide to business automation, then explore BPAI and an AI business process assessment.