Diligence · Operating decision guide
AI due diligence for private equity
Otomat editorial · Revised · Original edition 16 March 2026
An evidence request matrix for underwriting AI opportunity, data rights, operating readiness and downside before acquisition.
Underwrite an operating hypothesis
Begin with the investment thesis and one consequential workflow. Ask management to show the current process, source records and accepted output. A demonstration is useful discovery, but does not establish sustainable economics or authorization to use the underlying data.
Separate opportunity, dependency and exposure. A manual process may present an opportunity; unavailable rights, weak controls or scarce reviewers may make near-term execution unattractive. Record what was inspected, who owns it, its date and the remaining uncertainty.
Make the request actionable
Use the matrix below in a management or expert session. Select evidence the seller can provide lawfully; do not request production credentials or transfer sensitive client data to an unapproved model. Replace each blank with a record reference or “unknown.”
| Question | Evidence request | Owner / decision | Hold condition |
|---|---|---|---|
| Where is the constraint? | Task volume, cycle time, accepted outputs and exception sample for an agreed window | COO / workflow selection | No representative baseline |
| May this data be used? | Source contracts, retention terms, data classification and access approvals | Data owner + counsel / permitted scope | Rights or permitted destinations unknown |
| Can the system act safely? | Identity/action inventory; allow and deny tests; approval path | Security + technical owner / action boundary | Consequential action lacks independent approval |
| Can people operate it? | Reviewer availability; fresh-task and exception assessment | Functional manager / rollout cohort | No accountable reviewer or backup |
| Does the value bridge hold? | Full cost, counterfactual, cash mechanism and overlap register | CFO / scenario inclusion | Capacity claimed as cash without a release mechanism |
| Can we exit the supplier? | Export, restore and alternative-provider test; licenses and documentation | IT owner / transition cost | Required assets or rights unavailable |
Hold unsupported benefits outside the base case
Missing permission is a hold, not a low score to average away. An untested integration is a dependency with cost and timing uncertainty. An illustrative benefit belongs in a scenario, not accepted EBITDA. Ask finance to reconcile each proposed benefit to the value bridge and identify overlapping initiatives.
Carry the decision into the first operating mandate
Carry the diligence record into the operating plan: the selected workflow and investment hypothesis, a named management owner, baseline and source references, data-rights findings, unresolved dependencies, first acceptance test, full cost assumptions and stop rule. Recheck access, rights and management ownership after close. The implementation team accepts a scoped mandate; it does not inherit a forecast as an accepted result or delivery promise.
Use the working tools
Blank templates are ungated. Adapt them to the workflow; inputs are not submitted.
Sources and limits
These are Otomat’s proposed operating methods and original decision artifacts. External references inform risk management; they do not certify this method or establish client results. Linked sources were selected for this review; their live revision dates may change.
For anonymous documentation and proposal patterns, see delivery evidence and its boundaries.
Related decisions
Bring a workflow to the discussion
Share the business goal and constraints through the mandate process. Do not send sensitive production data in the initial brief.
Scope an Operating Mandate