Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Before investing in a lesser-known AI company, verify the issuer, customers, product performance, economics, data and intellectual-property rights, risks, capitalization, and actual investment terms. A polished demo, a financing announcement, a Form D filing, or a headline customer count is not a substitute for that evidence. The steps below are a general diligence process, with U.S. SEC filing examples; requirements vary by jurisdiction, industry, offering, and deployment.
1. Identify the company and the security you are considering
Start by establishing exactly which legal entity would receive your money and issue your security. Similar company names, a founder’s former business, a fund, a product brand, and the actual issuer can be easy to confuse.
- Record the issuer’s full legal name, jurisdiction of formation, subsidiaries, product or trading names, founders, and directors.
- Identify which entity signs customer and supplier contracts, owns the product’s intellectual property, employs the team, and issues the security. Differences between these entities may create important dependencies.
- Write down what security is being offered: for example, shares, a SAFE, a convertible note, or another instrument. Obtain the actual documents, not just a summary or pitch deck.
- Ask for a precise account of the customer, problem, workflow, product, and which part of the workflow AI performs. Separate what is live, in a pilot, on the roadmap, or only shown in a demonstration.
For each material company claim, request dated, underlying evidence. Mark what you have independently checked, what another source corroborates, what management has provided, what you have inferred, and what remains unresolved.
Recommended Free Tools
2. Check public records, but do not mistake a filing for approval
For a U.S. issuer or offering, search SEC EDGAR using the exact issuer name and any known CIK. If the company has made an offering that requires it, examine its Form D notices and amendments. Compare the issuer name, related persons, offering information, and filing dates with the documents the company provides. The SEC explains Form D filing requirements in its Form D FAQ and filing instructions.
#1 Best Overall
The SEC says a Form D notice is generally due within 15 calendar days after the first sale for specified exempt offerings. For this purpose, the first sale is when the first investor is irrevocably contractually committed. Form D filings are publicly available through EDGAR. The filing is a notice, not SEC approval, an audited financial statement, a complete capitalization table, or a guarantee that the offering is legitimate. The SEC staff FAQ itself says it reflects staff views and has no legal force or effect.
Depending on the company’s jurisdiction and market, relevant checks may also include corporate, court, patent, procurement, and regulatory records. A lack of public records does not establish that the company has no obligations, disputes, or operating business.
3. Establish whether customers pay, use, and renew
Ask for a customer list separated into paid production, paid pilots, unpaid pilots, and prospects. With the company’s permission, speak directly with a representative sample of current and former customers. Ask what they deployed, what it replaced, who approved the purchase, how often it is used, what measurable outcome changed, and whether renewal or expansion is expected.
Free tools Windows power users keep installed
One-click scans. No signup required.
Reconcile reported revenue with signed contracts, invoices, collections, credits, churn, and customer concentration. Review customer cohorts, retention and expansion, implementation time, and backlog conversion. Distinguish recurring subscription revenue from one-time services and integration work.
Do not treat a letter of intent, waitlist, benchmark result, demo, or pilot as equivalent to recurring paid use. If management reports “customers” or “AI users,” ask how those terms are defined, what activity counts, and which reporting period the figure covers.
4. Evaluate the product beyond its best demonstration
Arrange a demonstration, but define evaluation tasks independently instead of relying only on a vendor-selected showcase. Include representative inputs, edge cases, and failure-prone or adversarial examples, and compare the product with a conventional baseline or incumbent workflow.
Request the evaluation data and methodology, task-level error rates (and differences by user group where appropriate), human-review burden, latency, uptime, and evidence that results hold outside a curated demo. A company’s reported results are not independent verification; do not present them as tests you performed.
Map the production stack: foundation models, cloud and accelerator providers, retrieval or data vendors, open-source components, and human support. Ask for cost per completed customer task at observed and stressed usage, gross margin after inference and support, capacity commitments, rate limits, exposure to supplier price changes, and a fallback plan if a critical provider changes terms or withdraws access. A product built on third-party models can still be a viable business; the relevant questions are its dependencies, costs, and defensibility compared with a proprietary-model approach.
NIST’s AI Risk Management Framework is voluntary and intended to support trustworthiness considerations in AI design, development, use, and evaluation. NIST lists AI RMF 1.0 as released in 2023, a generative AI profile as released in 2024, and says the framework is being revised. It can provide a vocabulary for risk questions, but it does not certify a company or establish product quality, compliance, or investment merit.
5. Verify data rights, intellectual property, and security
Request a data inventory that covers training, fine-tuning, evaluation, retrieval, and inference. For each source, establish who collected it, what contractual or legal permission supports its use, what restrictions apply, whether it contains personal or confidential information, whether customers can opt out, and whether submitted data is retained or used to train shared models.
Review model and dataset licenses, employee and contractor invention assignments, third-party code and model obligations, patent and trademark claims, trade-secret controls, and any disputes or notices. A pitch-deck statement that data or models are “proprietary” does not establish ownership or permission; review the underlying records with qualified counsel.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Assess security architecture, access controls, encryption, logging, incident response, vulnerability management, and customer security commitments. Ask for incident history and remediation. For any independent audit or certification, check its date, scope, exceptions, covered systems, and the legal entity it covers. Review contracts for how responsibility is allocated if the system produces a harmful or materially incorrect result.
Best Value
6. Map governance, regulation, and exposure
Identify the markets where the system is offered and the consequential decisions it informs. With counsel familiar with those markets, determine which privacy, data-protection, consumer-protection, employment, health, financial, safety, export, and sector-specific rules may apply. Ask who is accountable for model changes, evaluation, incident escalation, customer representations, and board oversight. Review litigation, complaints, regulatory inquiries, insurance, indemnities, and contractual restrictions.
A SEC Investor Advisory Committee Disclosure Subcommittee document dated November 18, 2025, was a draft for discussion at a December 4, 2025, meeting. It recommends that the SEC consider issuer definitions of AI, disclosure of board oversight, and separate discussion of material AI effects on internal operations and consumer-facing matters. This is a committee recommendation draft, not an adopted SEC rule or a legal requirement for a private startup. It can inform questions about materiality and governance, but it is not a compliance checklist.
7. Reconstruct ownership and understand the investment terms
Obtain the current fully diluted capitalization table and reconcile it with the stock ledger, charter, board approvals, options, warrants, SAFEs, convertible notes, debt, liens, and prior financing documents. Identify promised equity and any side letters. Review liquidation preferences, anti-dilution provisions, conversion caps or discounts, information and voting rights, transfer restrictions, and follow-on obligations.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchRead the actual subscription, stock-purchase, SAFE, note, or other security documents. Model ownership and proceeds across multiple financing and exit outcomes, including dilution and downside. Compare valuations only after comparing the rights attached to the securities and the assumptions behind the valuation. Qualified legal and tax advisers can review the documents and your eligibility and jurisdiction; do not infer that an SEC notice has checked the offering terms.
8. Turn the evidence into an investment decision
Write a decision memo that states the thesis and downside case, and records customer proof, product evidence, unit economics, defensibility, key dependencies, governance and legal exposure, capitalization, and investment terms. For each conclusion, label the evidence as independently verified, corroborated, management-provided, inferred, or unresolved. Name the evidence that would change your view and specify any conditions that must be met before funding, such as customer verification, documented rights, security remediation, or clarified financing terms.
When comparing actual opportunities, keep the trade-offs visible rather than compressing them into a single “AI moat” score:
Quick Recap
| Compare | Evidence to examine |
|---|---|
| Customer demand | Urgency of the problem, willingness to pay, paid use, retention, and expansion. |
| Product delivery | Performance on representative tasks, implementation burden, and human-review needs. |
| Economics | Gross margin, inference and support costs, compute exposure, and supplier concentration. |
| Defensibility | Data and IP position, distribution, customer retention, and dependence on third parties. |
| Risk and financing | Governance and regulatory exposure, cash runway, valuation, dilution, and security rights. |
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
