Executive Summary
AI spending is climbing fast. Few companies can show it in the P&L.
U.S. private firms are on track to spend roughly $280 billion on AI in 2026, and per-employee spending rose about 50% in a single year (Federal Reserve Bank of Atlanta). The capital is committed. Whether it reaches the P&L is a separate question, and for private equity it is the one that matters.
A portfolio company can have six AI pilots, capable technical leaders, and a growing software budget, and still have no executive accountable for the financial result, no shared baseline, and no plan to convert saved time into earnings. The technology works. The value-creation system does not.
Traditional technology diligence establishes what systems, data, and models a company has. It rarely answers the more consequential question: can this management team and operating model convert those assets into measurable performance within the hold period? Octant calls that answer AI performance diligence.
Eight dimensions decide whether AI performs
The organizational lens of the Octant AI Reality Index℠, the AIR-O, scores eight connected dimensions. Octant advisors score each one from evidence, not from how the team rates itself. The output is not a maturity label. It is the two or three binding constraints, in the order they must be addressed.
| Dimension | Diligence question |
|---|---|
| Strategic Mandate | Is AI tied to a funded enterprise ambition and the deal thesis? |
| Executive Alignment | Do leaders agree on outcomes, priorities, ownership, and tradeoffs? |
| Governance & Decision Rights | Are AI decision rights clear, accountability defined, and risk actively managed? |
| Data & Technology Strategy | Is data governed as an enterprise asset, with clear build, buy, and partner decisions? |
| Operating Model & Structure | Can business and technical teams move from use case to changed workflow? |
| Leadership Capability | Can accountable leaders carry the transformation under pressure? |
| Workforce & Culture Readiness | Will roles, incentives, skills, and trust support adoption? |
| Value Measurement & Improvement | Are baselines, KPIs, benefits, and reinvestment decisions explicit? |
Two scores, and the gap between them
Every dimension gets two scores: a Claimed score, what leaders report, and an Evidenced score, what artifacts and observed decisions support. The Evidenced score is the one that counts. The distance between them is the Reality Gap, and a wide gap is often the most important finding, because the organization believes work is done that the evidence cannot confirm. Scores come from interviews, document review, and a station-by-station trace of the company’s own AI initiatives.
Used across the investment lifecycle
AI performance diligence runs at three decision points: underwriting capability and exposure before acquisition, converting the thesis into a sequenced agenda in the first 100 days, and turning exit claims into auditable proof across the hold. Every recommendation ships as a measurement contract, with a baseline, target, named owner, leading indicator, and test date.
Topics
The full framework covers the eight dimensions, the value gate, and the portfolio playbook, with a twelve-question field checklist. Free, no gate.