Before investing in an AI company, verify three things: customers get measurable value, the product performs reliably in real use, and the company can earn money on terms that make sense for you. Ask for evidence—not just a convincing demo or forecast—and assess the investment itself separately from the technology. The same diligence framework works for public shares and private offerings, but the documents and risks you can inspect differ.
Start with the investment you are considering
First establish whether you are evaluating publicly traded shares or a private offering. Public-company filings provide a recurring source of financial and risk information, though they do not prove that a product or forecast will succeed. A private offering may disclose less, and its terms can sharply limit your ability to sell or obtain information later.
| Investment | What to examine | Key limitation |
|---|---|---|
| Public shares | Latest filings, financial statements, share structure, risk factors, cash needs and potential dilution. | Public disclosure does not remove business risk or guarantee the shares can be sold at a favorable price. |
| Private offering | Offering documents, the security being sold, valuation basis, fees, use of proceeds, investor rights, transfer restrictions and the offering exemption. | Private placements can be illiquid, provide less disclosure and result in a total loss. Investor.gov says a Form D is a notice filing, not SEC approval or registration. See the SEC’s Private Placements under Regulation D – Updated Investor Bulletin. |
Use the questions below for either type of investment, then apply the relevant document checks. For a private placement, Investor.gov advises that if an issuer fails to answer questions adequately, investors should consider that a warning—not a reason to rely on assurances instead.
What customer problem does the product solve, and who pays?
Ask the company to identify the buyer, the user, the problem being solved and the budget that pays for it. A product announcement, free trial or pilot is not the same as a customer that has deployed the product, pays for it and renews.
#1 Best Overall
- How many customers are paying for the product, and how many are still in pilots or trials?
- What makes customers renew, expand their use or stop using it?
- How long does it take a customer to put the product into use and see value?
- What measurable result does the customer get, and how is that result calculated?
- Can customers confirm the claimed result, deployment status and renewal behavior?
Look for customer evidence that distinguishes durable use from a one-off experiment. Ask whether reported results are typical or drawn from a small number of selected customers, and whether the company can explain customer concentration. A large contract with one buyer may be less resilient than a broad base of renewing customers.
What does the AI actually do, and how well does it work in its intended setting?
Find out which tasks depend on AI, which parts are conventional software or human work, and what happens when the system is wrong. A polished demonstration shows a possible output; it does not by itself establish reliable performance in varied customer workflows.
- What evaluation or benchmark supports the performance claim, and does it resemble real customer tasks and data?
- How often does the system produce incorrect, misleading or incomplete outputs in the intended use?
- Which outputs require human review, and who is responsible for that review?
- How are failures reported, investigated and corrected?
- How does the company detect performance changes after deployment, including after models, data or customer workflows change?
Ask to see limitations and failure handling, not only favorable examples. Consider the consequences of errors in this particular use case: an incorrect suggestion for a low-stakes workflow is different from a flawed output that can affect a person’s health, finances, employment or legal position.
Does the company have the rights and controls it needs for its data?
Data can be an asset only if the company has the rights and operational controls to use it. Ask about data used to train or tune models, data used in the live product, and information customers provide. These may have different sources, permissions and retention rules.
Free tools Windows power users keep installed
One-click scans. No signup required.
- Where does each important dataset come from, and what rights permit collection, training, retention and commercial use?
- Can customer data be used to train or improve models, and what do customer contracts say about that?
- How is customer information separated, protected, retained and deleted?
- What are the company’s security practices and incident-response procedures?
- Which third-party models, data providers or infrastructure services are essential, and what contractual protections apply?
Press for specifics on privacy, security and intellectual-property rights rather than accepting a broad statement that data is proprietary or secure. If a critical model or infrastructure provider changes its price, terms or availability, ask how quickly the company could adapt and what that would cost.
Can the business make money as usage grows?
Revenue growth alone does not show that an AI product has attractive economics. Examine reliable financial statements where available and ask whether serving each additional customer—or handling more usage by an existing customer—improves or weakens the business.
- What are gross margins after model inference, computing, storage and other infrastructure costs?
- How do those costs change with customer usage, and can the company pass them on through pricing?
- What are customer acquisition costs, renewal and expansion patterns, and customer concentration?
- How much cash is being used, how long can current funds support operations, and what financing may be needed?
- What spending is required for research, infrastructure, security, compliance and customer support?
- Which assumptions connect the company’s forecasts to evidence from current customers and actual costs?
For a private offering, review the financial information provided and the proposed use of proceeds. Investor.gov’s private-placement guidance recommends assessing whether claims and expectations are reasonable as well as understanding how proceeds will be used. Treat projections as assumptions to test, not as established results.
What is defensible, and what could make the product replaceable?
Compare the company with direct competitors, incumbent software, in-house solutions, open models and the vendors it relies on. An AI label or access to a well-known model is not, on its own, evidence of a durable advantage.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors- What would customers lose, or have to rebuild, if they switched to another product?
- Does the company have durable workflow access, distribution, service quality or unique data rights?
- Could an incumbent add a similar feature, or could a customer build an adequate substitute?
- Can the company change models or infrastructure providers without disrupting service or materially increasing costs?
- How might changes in model capability, pricing or availability alter the product’s value?
Ask the company to describe where it wins and loses, not just name its competitors. A defensible position should be grounded in customer behavior, lawful access to necessary data, or other advantages that remain useful as the underlying technology changes.
What legal, security and governance obligations apply?
Obligations depend on what the product does, where it is offered and who uses it. Ask which jurisdictions and customer industries matter, and whether the company has assessed the rules and contracts that apply to those uses.
- Who reviews consequential outputs, handles escalations and decides whether a system should be paused?
- Who is contractually responsible if an output causes harm, and what indemnities or insurance exist?
- How does the company address privacy, intellectual-property claims, security incidents and customer requirements?
- Who oversees model changes, evaluations, access controls and incident reporting?
- Does the company have qualified legal advice for the markets and use cases it serves?
Do not assume that a general compliance statement settles these questions. Have qualified counsel assess the specific company, use case, contracts and jurisdictions when the exposure is material to your decision.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What are you buying, and what can change your ownership or exit?
Separate the quality of the business from the rights attached to the investment. For public shares, understand the share structure and financing needs. For a private security, read the governing documents closely; a headline valuation does not tell you what the security entitles you to receive or when you might be able to sell it.
Best Value
- What security are you purchasing, and how was its price or valuation determined?
- What fees, liquidation preferences, conversion terms, voting rights and information rights apply?
- How could later fundraising, employee equity or other securities dilute your ownership?
- Are there transfer restrictions, and what realistic paths to liquidity exist?
- Could you afford to hold the investment indefinitely or lose the full amount?
Do not treat an anticipated acquisition, public listing or future financing as a guaranteed exit. For private placements, Investor.gov notes that the investment is highly illiquid compared with one purchased on a stock exchange.
Who is making the claims, and what incentives do they have?
Verify important claims independently and understand the incentives of founders, promoters, brokers and other representatives. Ask for management backgrounds and references, and check whether the people presenting the opportunity are compensated for a sale or have interests that differ from yours.
- Can management substantiate material claims about customers, performance, finances and data rights?
- Are answers direct, consistent with the documents and clear about uncertainty?
- Do representatives explain fees, conflicts and the risks as readily as the potential upside?
- Is there pressure to decide immediately, a promise of guaranteed returns, or a claim that the SEC has approved the offering?
Urgency, implausible claims, missing information and evasive answers are reasons to pause and verify. Investor.gov’s guidance on private placements identifies high-pressure pitches and conflicts as warning signs; a claim of SEC approval is especially misleading when a Form D is only a notice filing.
Compare companies using the same evidence standard
When weighing two or more AI companies, do not let a prominent investor, a particular model choice or an “AI-first” label substitute for comparable evidence. Apply the same questions to each company, and record what is documented, what management asserts and what remains unknown.
Recommended Free Tools
| Comparison area | Evidence to compare |
|---|---|
| Adoption | Paid deployments, renewals, expansion, customer concentration and customer-verified outcomes. |
| Product performance | Results in real workflows, known failure modes, human oversight and time to value. |
| Data and dependencies | Data and IP rights, privacy and security controls, and dependence on model or computing providers. |
| Economics | Gross margin after AI infrastructure costs, cash use, capital requirements and financing risk. |
| Durability | Customer switching costs, competitive alternatives and resilience to technology or supplier changes. |
| Exposure | Legal and governance issues associated with the company’s use cases and operating locations. |
| Investment terms | Valuation, security rights, dilution, fees, liquidity and plausible exit options. |
For private companies, information may be limited and valuation harder to assess than for listed shares. Make that uncertainty part of the comparison rather than filling gaps with assumptions.
Quick Recap
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.
