The October answer
An AI deployment matters in diligence only when its economics and controls can be verified. For each material agent, the portfolio company should be able to show its cost, identity, data access and measured workflow result.
What this edition adds
The September report covered sign-off, funding, headcount substitution, expected payback and access. This edition adds evidence on production-agent counts, cost visibility, operating bottlenecks and credential models.
Six diligence questions
- Who owns AI purchasing and who owns each production agent?
- Is the work funded with new money, reallocated software spend or money previously intended for headcount, and is there a clear budget at all?
- What share of AI operating cost is visible by workflow?
- Which agents use shared accounts or excessive permissions?
- What data bottleneck limits scale?
- Which measurable result supports the valuation case?
The value-creation read
Money that would otherwise have gone to headcount appears in 19 percent of finance-instrument answers (base 389, any mention). That does not, by itself, establish a labor saving: the workflow, cost and service level must also change. Separately, 13 percent of AI Leaders respondents describe AI as embedded enough that removing it would alter the cost structure or hiring plan (base 91).
The risk read
Agent access is the leading security concern at 68 percent, while 34 percent report a dedicated AI security budget line (base 151 for both, any mention). Shared service accounts are used by 37 percent of applicable AI Leaders respondents (base 49). In a transaction, those findings may affect remediation costs, integration plans and the evidence behind an EBITDA adjustment.
Private equity diligence matrix
| Diligence question | October evidence | Base | What to verify |
|---|---|---|---|
| Who can authorize AI? | CEO 49%; finance 31%; no single owner 13% | 467 | Sponsor, operating owner and approval record |
| Where does the money come from? | 41% net-new; 31% no clear budget; 19% headcount-linked | 389 | Recurring run cost, displaced spend and benefit category |
| What blocks scale? | 54% name proving ROI | 389 | Baseline, measurement window and decision threshold |
| Is production cost visible? | 44% full; 39% partial; 17% none | 77 | Workflow-level cost and gross-margin sensitivity |
| How do agents access data? | 37% shared accounts among applicable answers | 49 | Attribution, privilege, logging and revocation |
| Is security funding explicit? | 68% name agent access as a problem; 34% name a dedicated budget line | 151 | Control roadmap and funded remediation |
The matrix distinguishes a management claim from evidence that can support a deal model. Headcount-linked funding may represent cash savings, avoided hiring or redeployed capacity; those outcomes are not interchangeable. A production workflow with only partial cost visibility also cannot support a precise margin claim.
| Group | Base | Full visibility | Partial visibility | No visibility |
|---|---|---|---|---|
| Exploring | 18 | 22% | 33% | 44% |
| Piloting | 14 | 36% | 43% | 21% |
| Deployed in production | 33 | 61% | 33% | 6% |
| Embedded (removing it would change our cost structure or hiring plan) | 10 | 50% | 50% | 0% |
Matched respondents answering both questions; base 75. Small row bases are directional.
In the matched sample, full cost visibility is 22 percent among explorers and 61 percent among respondents with production deployments. Even so, 39 percent of the production group report only partial or no real-time visibility. The sample is selective and the row bases are small, so the result should guide diligence questions rather than valuation assumptions.
Comparison with external research
McKinsey's 2026 work on generative AI in private markets warns that AI-generated diligence can diverge from expert evidence on market size, pricing and margins. The same standard should apply to portfolio-company claims: reconcile them to the workflow, cost, identity and measured outcome.
What this means for private equity
Add an agent register and an AI cost ledger to operating diligence. Reconcile any claimed saving with headcount, software, infrastructure and control costs. “In production” identifies what to test; it is not proof of value.
Questions this report answers
What should private equity test in AI diligence?
Test ownership, funding source, workflow-level cost, agent identity, data dependencies and the measurable result behind each material deployment.
How common is unclear AI budgeting?
Thirty-one percent report no clear AI budget (base 389, any mention).
How do production-agent adoption and cost visibility compare?
Eighty percent report at least one production agent, while 44 percent report full real-time cost visibility: a 36-percentage-point difference across two questions from the same event (bases 75 and 77).
Key citable facts
- Thirty-one percent report no clear AI budget (base 389).
- The share reporting at least one production agent exceeds the share reporting full real-time cost visibility by 36 percentage points across two questions from the same event (bases 75 and 77).
- Agent access named as a problem exceeds the dedicated-budget-line share by 34 points (base 151, any mention).
Methodology and limitations
This is a synthesis of operator data, not a survey of private equity professionals. It does not estimate market size, company valuation or causal impact. Bases from separate instruments are reported independently. The 34-point security comparison is the difference between 68 percent naming agent access as a problem and 34 percent naming a dedicated AI security budget line; it does not measure budget adequacy, and respondents could report other funding sources.
Related reading
Citation
Suggested citation: Newlands, M. (2026). Private Equity AI Report, October 2026. Open Future Forum.
Work These Questions with Dealmaking Peers
Forum Select meets through small, off-the-record gatherings. Membership is by application and referral.