The IC Evidence Standard for Investing with AI

The IC Evidence Standard for Investing with AI

Investment committees do not need more confident prose. They need to know what supports a conclusion, how current the evidence is, which assumptions shaped the analysis, and what remains unresolved. As AI becomes part of research, diligence, and modelling, that standard matters more, not less.

The risk is easy to miss. An AI-assisted answer can sound polished while relying on an old filing, an ambiguous company identity, an unsupported adjustment, or a model that did not converge. By the time the answer reaches an IC memo, those conditions may have disappeared behind a neat paragraph and a clean chart.

Investment Committee Standards for AI-Assisted Deal Underwriting

Picture the moment before an investment committee meeting. A partner points to a downside case and asks, “What changed this number?” The analyst should not have to search a data room, a chat history, and three versions of a workbook to reconstruct the answer. The evidence chain should already be there: source, observation date, assumption, model run, warning, and review decision.

That is the IC evidence standard.It does not ask AI to be infallible. It asks the workflow to make its basis inspectable, especially when the available evidence is incomplete.

Every material claim should retain a source identity and enough context to retrieve it. Time-sensitive evidence should separate the date an event or value was observed from the date it entered the platform. If coverage is partial, the gap should be visible rather than turned into a silent estimate.

Every material numerical result should expose its assumptions, methodology and version, relevant warnings, and enough information for a reviewer to reproduce the analysis. Calculation or solver failure belongs beside the result, not in a technical appendix that is easily overlooked.

Review status matters too. A generated draft, an analyst-reviewed finding, and an approved IC assumption are not the same thing. A useful workflow makes those states obvious so a fluent draft is never mistaken for an authorised decision.

Creating an Investment Committee Evidence Register and Assumption Log

Consider a synthetic credit investment. A covenant-headroom output draws on a borrower schedule, a contract term, and an analyst adjustment. The register records when each source was observed, the transformation applied, the scenario assumptions, the model version, any warning, and the reviewer who accepted or rejected the adjustment.

If the committee challenges the downside revenue case, the team can trace the affected outputs and rerun the scenario with a different approved assumption. The old decision remains visible. Nothing depends on remembering which cell someone changed on the evening before the meeting.

The UK Government's 2025 AQuA Book recommends explicit data, assumption, and decision logs. Its assumptions log records what an assumption is, its effect and reliability, when and why it was made, and who made and approved it.[1]

An IC workflow needs the same separation. Evidence answers what the source says. An assumption records how the team interprets incomplete evidence. A decision records what the authorised reviewer accepted. Keeping these states distinct prevents a generated interpretation from quietly becoming an approved model input.

The Investment Committee AI Evidence Checklist

Before any AI-assisted recommendation is presented for partner vote or investment committee sign-off, the deal pack must satisfy four verification standards:

• 1. Source Provenance: Every historical figure, customer retention metric, and contract term is linked to a verified source filing, virtual data room document, or audited financial statement.
• 2. Disclosed Assumption Log: All forecast parameters, EBITDA adjustments, revenue growth trajectories, and exit multiple assumptions are explicitly isolated from reported facts.
• 3. Deterministic Solver Validation: Valuation and return metrics (IRR, MOIC, DSCR) are computed by peer-reviewed quantitative engines with mathematical convergence proof, not unverified text approximations.
• 4. Challenge & Replay Capability: Committee members can alter any single input during the meeting and instantly rerun the deterministic sensitivity model with full auditability.

ILPA Performance Standards and Private Equity Valuation Guidelines

ILPA developed its Performance Template to standardise return calculation methodologies and the capture of performance metrics with related contributions and distributions.[2]The template supports comparability and review. It does not make historical performance predictive. An IC record should apply the same discipline to definitions, while preserving uncertainty about future outcomes.

The IPEV Valuation Guidelines likewise set out current best practice for reporting private capital investments at fair value.[3]The committee should be able to see which valuation evidence and judgement produced a number, not only the number itself.

Not every statement needs the same evidence burden. A market description may need a current source and a clear limitation. An EBITDA adjustment that changes leverage or returns needs source reconciliation, calculation detail, ownership, and approval. A legal conclusion needs qualified review. The evidence standard should tighten as the claim becomes more consequential to the decision.

Resiliq connects research that preserves sources, governed diligence workstreams, quantitative services, and reviewable outputs. Quantitative results can carry assumptions, methodology provenance, warnings, and reproducibility information. Diligence workflows can show which workstreams completed and which evidence is still missing before synthesis.

Sources still require professional judgment about completeness, freshness, and correctness. Legal interpretation, accounting judgement, materiality, and final approval remain human responsibilities. The point is not to remove uncertainty.It is to stop uncertainty from disappearing on its way to the decision.

Validating Investment Memo Assumptions and Auditability Under Scrutiny

Choose a recurring decision such as a target-screen approval or a quality-of-earnings adjustment. Define the claims and numbers that require evidence, name the review states, and test whether the evidence chain survives the move from research to memo to model.

An evidence standard should work when the case moves, not only when the first memo is assembled. Change one material source, remove an adjustment, or revise a financing assumption. The record should show which findings, calculations, scenarios, and narrative conclusions require review.

This test separates a refreshed document from a refreshed decision. Updating a chart is not enough if the surrounding conclusion still depends on the old evidence.

Not every sentence in an IC paper needs the same treatment. Material claims, numerical drivers, risk conclusions, and approval conditions need stronger lineage and review than background context. The standard should become stricter when an output can change valuation, financing, structure, risk classification, or the decision to proceed.

Proportionate assurance keeps the process usable. It directs reviewer attention to the claims that matter most without turning the entire paper into an audit file.

Build your next IC case in Resiliq and test whether every material claim, assumption, warning, and open question survives senior review.

Important notice

This article provides general information only. It is not investment, valuation, financial, legal, tax, accounting, financing, or other professional advice, recommendation, solicitation, or offer concerning any company, security, transaction, strategy, or product. Examples are illustrative and not forecasts. Resiliq references describe capabilities reviewed at the time of writing, not promises of future availability, performance, or outcomes.

References

1. UK Government Analysis Function, The AQuA Book, 2025
2. ILPA, Performance Template, version 1.1, 2025
3. International Private Equity and Venture Capital Valuation Board, IPEV Valuation Guidelines 2025