What Private Opportunity Screening Actually Means
Private opportunity screening is the process of deciding whether a confidential company, investment project, or transaction deserves a closer look before broader circulation. For an AI private deal-flow network serving founders and operators, it is less about judging whether an opportunity is exciting and more about testing whether the information is credible, relevant, actionable, and compatible with the network’s mandate. As of 25 September 2026, this function has become more structured because AI-related opportunities now span software, infrastructure, semiconductors, data services, cybersecurity, fintech, healthcare, robotics, and energy—not merely companies that market themselves as “AI startups.” A good screening system separates these categories before discussing valuation or deal size.
Also worth reading: Can Private AI Dealmaking Networks Unlock More Founder Opportunities? · What makes an AI deal sourcing platform for founders different from traditional networks? · How do ai deal sourcing tools compare for private equity?
The term should not be confused with airport security screening, public-health screening, or investment-screening tools unrelated to private markets. In a deal-flow setting, “screening” means an initial gate, not a final investment decision. A compact screening process might review the company’s stage, use of proceeds, target capital, revenue evidence, customer concentration, technical dependency, regulatory exposure, and fit with participating members. The key word is “private”: unlike a public stock, the underlying company and most financial information may not be available for independent verification. That information asymmetry makes disciplined screening especially important.
A responsible network should preserve confidentiality, document where each claim originated, and distinguish verified facts from management representations. It should also state what it cannot know. The objective is not to reject uncertainty; private companies naturally contain uncertainty. The objective is to identify which uncertainties matter, whether they can be reduced through diligence, and whether the opportunity has enough potential to justify the next stage. Screening produces a ranked or categorized outcome rather than an automatic yes-or-no verdict.
Why AI Deal-Flow Networks Need Their Own Screening Method
AI opportunities create unusually broad gaps between presentation quality and economic substance. A polished demonstration can hide a weak data-rights position, an expensive inference burden, a dependency on one model provider, or a sales process built around unpaid pilots. At the other extreme, a technically modest product with recurring revenue and a defensible distribution channel may be more attractive than an ambitious foundation-model project. Generic startup scoring often misses these differences because it treats sector, capital intensity, sales cycle, technical risk, and governance requirements as interchangeable.
A useful private-market framework separates four questions: Is the opportunity real, is it investable or otherwise actionable, is it appropriate for this network, and is the timing acceptable? Reality covers the legal existence of the company, product operation, customer relationships, and provenance of supplied metrics. Investability covers economics, financing needs, competitive durability, and the feasibility of the requested capital. Fit concerns the network’s members, geographic mandate, risk preferences, and ability to assist. Timing covers financing conditions, technology readiness, regulatory deadlines, and whether delay could impair the opportunity.
Screening also protects members from false precision. A platform may ask for a 12% projected return, but an unverified forecast is not comparable to audited revenue. Likewise, “100 customers” means little unless the reader knows whether they are paying customers, whether contracts are annual or monthly, and whether a single customer represents 40% of revenue. The best network records evidence quality, not just the most favorable number. As of 2026, AI systems can organize documents, extract claims, and flag inconsistencies, but the accountable decision must remain with a named human reviewer.
A Six-Step Process for Testing an Opportunity
The first step is intake and classification. The submitter should provide the company name, jurisdiction, operating history, stage, product category, capital target, use of proceeds, expected runway, and the reason the opportunity is being offered. The network should verify basic identity and identify whether the submission is a company, fund, SPV, acquisition target, commercial partnership, debt request, or secondary transaction. This takes little time for complete submissions and much longer for vague ones; incomplete intake should not be treated as a negative judgment, but rather as an unresolved diligence item.
The second step is evidence mapping. Every important assertion should be linked to a source, such as a cap table, bank statement, customer contract, financial model, product log, patent record, or management response. The reviewer should label evidence as independently verifiable, third-party supplied, management-supplied, projected, or unknown. The third step is an initial economics check, including current revenue, recurring versus nonrecurring revenue, gross margin direction, monthly burn, cash balance, and expected runway after financing. A rough threshold is six months of additional visibility, although 12 to 18 months is often more appropriate for a network that expects a conventional funding process.
The fourth step tests the AI-specific value chain. Reviewers should determine whether the company owns proprietary data, builds a defensible model, controls distribution, provides workflow software, supplies compute, or merely resells another provider’s technology. They should examine reliance on external APIs, compute providers, hardware suppliers, and regulated data sources. The fifth step is a conflict and confidentiality check, including whether the submitter is entitled to circulate the materials and whether one member’s participation would compromise another negotiation. The final step assigns a category such as proceed, reserve, decline, or insufficient information, with reasons and the next action recorded.
What the Screening Committee Should Measure
A screening committee should use a small number of comparable measures rather than an elaborate score that creates the appearance of accuracy. Commercial evidence can include annual recurring revenue, net revenue retention, customer concentration, pilot conversion, gross margin, and contracted backlog. Technical review can include model dependence, reproducibility, data rights, evaluation quality, security controls, and inference cost. Governance review should cover cap-table cleanliness, founder alignment, related-party transactions, intellectual-property ownership, and prior financing claims.
Numbers should be normalized before comparison. A company reporting annual revenue should not be directly compared with one reporting signed contract value, and a valuation based on the next twelve months of projected revenue is not equivalent to one based on current revenue. Useful reference points might be a customer representing more than 25% of revenue, runway below nine months after the proposed raise, or a valuation requiring more than three times current recurring revenue without a documented expansion path. These are review triggers, not universal pass-fail rules. Enterprise infrastructure businesses can rationally sustain lower near-term margins than low-cost software companies.
A practical scoring sheet can weight commercial evidence at 30%, technical defensibility at 25%, unit economics or capital efficiency at 20%, management and governance at 15%, and fit plus timing at 10%. The committee can then impose hard gates for legal identity, data rights, sanctions compliance, security, and confidentiality. The weights should not be manipulated to reach a predetermined result. Their purpose is to make disagreements visible: if two reviewers reach opposite conclusions, the scoring record should show whether the cause is different evidence, different assumptions, or different priorities.
| Feature | Basic screen | Committee screen | Full investment diligence |
|---|---|---|---|
| Typical timing | 24–72 hours | 1–3 weeks | 4–12 weeks or longer |
| Inputs | Submission, stage, sector, capital target | Verified financial, customer, technical, and governance evidence | Extensive legal, technical, commercial, tax, and financial work |
| Decision | Route, reserve, or request more information | Proceed to a named next stage or decline | Investment, term-sheet, or transaction recommendation |
| Financial depth | Headline revenue and runway | Monthly cash flow, concentration, margins, and normalized metrics | Audited or otherwise deeply tested records and detailed forecasts |
| AI-specific review | Product and infrastructure classification | Data rights, model dependence, compute needs, and security flags | Code, model, architecture, data provenance, and customer-use verification |
| Expected accuracy | Directional only | Decision-useful but still incomplete | Highest feasible confidence under normal diligence limits |
Private opportunity screening itself can be inexpensive, but the cost depends on depth and source. An internal network may perform an initial screen with existing staff and AI-assisted document tools, limiting direct expense to staff time and subscriptions. A specialist service may charge an hourly fee, a fixed project fee, a success fee, or a combination. As of 2026, prices vary too widely for an honest universal range, but a lightweight commercial and governance review might cost several thousand dollars, while technical, legal, and financial diligence can move into tens of thousands or substantially more.
Members should distinguish among three separate charges. Screening is the fee for deciding whether to continue. Transaction support, such as negotiation, data-room management, or closing coordination, should be quoted separately. Fundraising advisory or capital-raising fees may also apply and should be described transparently, including any relationship to the target or intermediary. A free listing sounds attractive, but it can be expensive if it lacks a real verification process; conversely, an expensive review does not guarantee a sound investment.
The total financing need should include post-raise runway, not merely the headline capital gap. If a company has $1.5 million cash, burns $250,000 per month, and raises $5 million, simple runway extends from six months to 26 months. If fundraising and deal costs consume 12%, the actual new cash is $4.4 million, producing about 23.6 months of runway. Reviewers should also test whether a model assumes rapid hiring, lower inference prices, or immediate pilot conversion, because even a 20% variance in burn can erase several months of visibility.
A network may also need to disclose who pays and who receives the economics. If a deal-screening firm is compensated only when a transaction closes, its incentive may favor more circulation or stronger promotion than neutral screening. Independent fixed-fee review can reduce this conflict, while success fees should be paired with an agreed screening standard and independent oversight. Any compensation tied to investor participation should be visible to the submitting company and relevant members.
Alternatives to Centralized Screening
Founders do not always need a network-level committee. A founder seeking capital or a strategic partner can use an experienced advisor, an accelerator, an institutional venture team, an angel syndicate, or an industry-specific data room. Each alternative has different biases. An accelerator may provide useful coaching but be limited in follow-on resources. An institutional fund may conduct stronger commercial diligence but normally serves companies matching its fund mandate. An operator may notice customer and execution risks that a software reviewer misses, yet may have less time for legal and financial checks.
Self-screening is appropriate for an early internal decision about whether to continue preparing a company for outside circulation. The founder can build a 10 to 20 page screening memo, attach source documents, and ask two independent reviewers to challenge it. This approach costs little and is useful before claiming that materials are investor-ready. It is not a substitute for independent diligence because founders are both promoters and evidence sources. The same principle applies to AI networks: automation can locate contradictions, but it cannot eliminate the conflict created by compiling a company’s own materials.
| Screening route | Best use | Main advantage | Main limitation |
|---|---|---|---|
| AI-assisted internal review | Sorting and first-pass inconsistency checks | Fast and inexpensive at scale | Depends on correct source material and human oversight |
| Experienced operator review | Commercial, product, and execution assessment | Understands real operating constraints | May lack financial or legal capacity |
| Specialized advisory firm | Structured commercial or technical diligence | Repeatable procedures and relevant expertise | Adds fees and can still depend on company cooperation |
| Accelerator or fund process | Ecosystem support and capital-readiness review | May connect the company with a broader financing network | Fit and capacity are limited by that institution |
| Full diligence team | Legal, financial, tax, security, and technical verification | Highest practical evidence standard | Slow, expensive, and not free from uncertainty |
A major mistake is treating sector excitement as evidence. The growth of generative AI does not validate every model, wrapper, consultancy, or compute claim. Another error is counting pilots as customers. Reviewers should ask for paid invoices, contract duration, renewal behavior, deployment scope, and the identity of the economic buyer. A pilot count should never be added to recurring revenue without an explicit estimate of conversion probability, and that estimate should remain separate from realized revenue.
Confidentiality is also mishandled. Circulating a complete data room before identity, rights, and audience are confirmed may expose a founder to unauthorized outreach, copycat behavior, or loss of negotiating leverage. Networks should use access tiers, watermarking where appropriate, expiring links, and a documented distribution list. Material nonpublic information rules may apply when a company is public or when transaction information originates from a public issuer, so legal counsel may be needed even when the target itself is private.
The process should avoid false binaries and excessive metrics. A score of 82 does not mean there is an 82% probability of success, and a long questionnaire does not compensate for poor source evidence. Reviewers frequently overlook related-party loans, inconsistent share counts, intellectual property assigned to a former employer, customer contracts terminable at will, or planned revenue built on unowned data. The safest review converts open questions into assigned follow-ups with owners and deadlines.
When to Act, Pause, or Decline an Opportunity
A network should move quickly when basic identity is verified, the capital target is plausible, the company has a defined use of proceeds, the founder provides quality source documents, and no material conflict appears. A reasonable target is to return an initial routing decision within two to three business days for complete submissions, with a committee decision within 10 to 15 business days if the opportunity warrants deeper review. Those service levels should not be promised for complex cross-border, regulated, or capital-intensive transactions.
Pause when customer concentration is extreme, runway is already exhausted, ownership documents conflict, or essential product claims cannot be tested. A useful trigger is insufficient information affecting more than 20% of the modeled enterprise value. A pause should specify what is missing—for example, the top 20 customer contracts, monthly bank statements for six months, the full cap table, or proof of data licensing. It should not merely say “await further diligence.”
Decline or close the file when founders will not address basic legal, ethical, security, or regulatory questions; when the requested economics depends on undisclosed related-party arrangements; when materials appear misleading; or when the opportunity conflicts with network rules. There is also no duty to maximize the number of deals. A network with 20 well-screened opportunities may be more useful than one with 200 loosely classified submissions, particularly when members’ time and trust are the scarce resource.
The final recommendation should be conditional and time-bound. It might state that the company can enter a commercial-review stage if it supplies six months of bank statements and clarifies a 31% customer-concentration figure by a specified date. This protects momentum without confusing interest with approval. For the Mercer Club NYC, the non-promotional role is to organize this process: define the network’s mandate, verify submissions, record evidence, surface conflicts, and make the decision path legible to founders and operators. Private opportunity screening should increase the quality of conversations, not manufacture confidence where the facts do not support it.