A co-investor intelligence protocol for uncertain AI deals

Use network evidence without mistaking activity for success

Hyun and Kim analyze historical venture syndication networks and the number of AI-startup investments made by Korean and Japanese VC firms. Their measures of reachability and brokerage help explain investment activity under uncertainty. The outcome is investment count, not return, technical validity or social benefit. A family office cannot infer that a connected sponsor, frequent investor or brokered introduction identifies a superior deal.

The protocol treats a network as a source-discovery map. It asks what knowledge may travel through a relationship, how current it is and whether it can be independently checked. Reputation, warm introductions and co-investor participation are recorded as context, never as substitutes for evidence.

Map the deal network and interests

The deal lead records founders, directors, current and proposed investors, advisers, vendors, major customers, academic affiliations and prior co-investments. For every referral, the file identifies the referrer’s relationship, compensation, ownership, board role and access to nonpublic information. Consent and confidentiality boundaries are documented before references are contacted.

The map distinguishes direct experience from repeated hearsay. It flags circular validation in which several parties rely on the same unverified claim. Compliance reviews placement fees, political exposure, sanctions, confidentiality and material nonpublic information under current law. A network score is not created from protected identity or personal status.

Verify technical claims independently

A qualified technical reviewer defines the claimed AI capability, model inputs and outputs, evaluation population, baseline, failure modes, data rights, security architecture, compute dependencies and human oversight. The startup supplies reproducible evidence within lawful disclosure limits. Demonstrations are separated from production performance.

The review tests representative cases, edge conditions and unsupported-output rates. It checks whether reported gains come from the model, workflow changes, human review or selective samples. If evidence cannot leave a secure environment, the reviewer records the procedure and limitations rather than converting a founder statement into a verified result.

Test the business and financing case

Commercial diligence identifies the buyer, budget owner, procurement cycle, integration burden, retention, pricing, gross margin, service labor and regulatory constraints. Customer references distinguish pilots, paid deployments and renewals. Finance reconciles revenue, cash, burn, runway, capital needs, preference terms, dilution and follow-on assumptions.

The committee models failure to raise the next round, slower adoption, higher compute or compliance cost, customer concentration and model commoditization. Co-investor reserves are not counted unless legally committed. A strong network cannot cure an unaffordable valuation, weak rights or a product without a defensible customer need.

Govern syndication and information sharing

The investment committee defines the lead investor’s duties, family-office information rights, allocation, pro rata rights, board observation, conflicts, side letters and decision deadlines. Counsel reviews whether investors receive unequal material information and how confidentiality survives if the office declines. The office retains its own investment memo and does not outsource fiduciary judgment to the syndicate.

Post-close communication has a named owner and source record. Introductions to portfolio companies, customers or experts require authorization and a stated purpose. The office tracks whether network support actually occurs instead of counting promised access at underwriting.

Measures and stopping conditions

Measures include claims independently reproduced, unresolved technical exceptions, paid-customer evidence, runway under downside cases, concentration, reference conflicts, committed follow-on capital, information-right compliance and value delivered by promised network support. Reviews occur after material model, data, regulatory, financing or leadership changes.

The office stops or pauses when technical claims cannot be reproduced, data rights are unclear, the referral chain is circular, conflicts are undisclosed, customer evidence is misrepresented, the syndicate pressures a deadline before required diligence, or downside financing exceeds approved capacity. It records the reason and preserves contrary evidence. Neither the source nor this protocol predicts returns or validates an AI company.

Learn from declined and failed opportunities

The research lead maintains a decision journal for deals approved, declined and abandoned. It records which claims later proved true, which risks emerged, and whether network introductions produced information unavailable through ordinary diligence. The review uses only lawfully obtained data and avoids rewriting the original rationale with hindsight.

Quarterly calibration compares predicted milestones with observed outcomes and checks whether familiar sponsors receive lighter diligence. Repeated optimism, missing downside cases or reliance on the same expert network triggers a process review. Findings update checklists and evaluation sets, but they do not retroactively convert a poor outcome into negligence or a good outcome into proof of sound process.

Source, attribution and limits

[1] Eun-jung Hyun; Brian Tae-Seok Kim. Overcoming Uncertainty in Novel Technologies: The Role of Venture Capital Syndication Networks in Artificial Intelligence (AI) Startup Investments in Korea and Japan. Systems (2024-02-24). https://doi.org/10.3390/systems12030072. CC BY 4.0.

The study analyzes investment counts for Korean and Japanese VCs using historical network measures and negative-binomial models. It does not study family offices, causal network interventions, investment returns, startup outcomes or the quality of AI claims. Network position may reflect selection and omitted factors.

Prepared by the Family Office Private Wealth Journal editorial research desk with AI assistance at the direction of Dr. Freedom Cheteni. Original research authors retain full credit and are not authors or endorsers of this companion. This proposed process has not been evaluated in a new empirical study; no independent peer-review acceptance is claimed. This is educational analysis, not individualized investment, legal or tax advice.

Implementation record

Open the cited sources

1. [1] Eun-jung Hyun; Brian Tae-Seok Kim. Overcoming Uncertainty in Novel Technologies: The Role of Venture Capital Syndication Networks in Artificial Intelligence (AI) Startup Investments in Korea and Japan. Systems (2024-02-24). ↗