Beyond the AI headline: what does the business sell?

“It is an AI company” can describe businesses with very little in common. One sells computing hardware. Another rents access to computing capacity. A third sells software that uses that capacity. Another combines software with machines operating in the physical world.

For a reader who understands operating businesses, this is an advantage: familiar questions about customers, costs, service and repeat purchases remain useful. You do not need to predict the future of artificial intelligence to ask what a particular company sells today. Nor does understanding a product make its security an appropriate investment.

The distinctions below are an editorial framework, not a forecast about which companies or categories will win. They make no market-size claims and do not imply that a technology category produces a particular return.

Find the customer and the budget

Who pays, who uses the product and who approves the purchase? They may be different people. An employee enjoying a demonstration is not the same as a business committing a budget. A trial, a letter expressing interest and recurring paid use are different forms of evidence.

Ask what task the customer needs done and what they use instead. The alternative might be another software product, a contractor, existing equipment or doing nothing. “Better technology” is incomplete unless it changes something the customer cares enough to purchase.

Then ask where the spending comes from. Does the product replace an existing expense or require a new budget? Does adopting it require procurement, security review, training or changes to a workflow? Those frictions matter even when the demonstration is impressive.

Distinguish four kinds of business

These are starting points, not rigid boxes. A company can operate across layers. The useful exercise is to identify which activities generate revenue and which consume cash, rather than classifying everything under the same headline.

This conceptual diagram separates business activities; it does not rank companies or predict the movement of investment returns.

Follow the cost of serving a customer

Ask what happens when usage grows. Does each additional task require paid computation? Is human checking part of delivery? Are installation and ongoing support included in the price? Revenue growth can look attractive while the cost of delivering the service grows alongside it.

For software, request an explanation of revenue after direct delivery costs rather than assuming software-like economics from the name. For physical systems, ask about equipment, deployment and service work. For hardware, ask how cash moves through purchasing, inventory and payment collection.

Do not confuse a company's current economics with a hoped-for future improvement. “Costs may fall” is a claim to examine, not a present margin. “Customers will use more” needs evidence about behavior, not just enthusiasm for a category.

Test usefulness beyond the demonstration

What can go wrong in ordinary use? Ask how the product handles errors, data access, reliability and human oversight. Which work can it do consistently, and which work still requires a person? An attractive demonstration can be deliberately narrow; a customer has to live with the full workflow.

Request evidence that matches the claim. Paying customers support a different conclusion from website visits. Repeat use supports a different conclusion from registrations. Ask whether references, contracts or financial information can be examined and what limitations remain.

Investor.gov's private-placement bulletin specifically encourages investors to question reliance on technology claims, competitors, management, financial statements and the planned use of money. Its broader warning is limited disclosure: the material in a private offering may not answer every question you need answered. Investor.gov: private placements.

Bring five questions to the next family call

• Who pays for the product, and what problem are they paying to solve?
• What exactly is delivered: equipment, capacity, software or a service?
• What does serving a customer cost, including people and support?
• What evidence shows continued paid use rather than initial curiosity?
• Which assumptions must become true for the business to sustain itself?

Write the answers in ordinary language. If a sentence depends on a technical term, ask what that term changes for the customer or the economics. If the answer is unknown, label it unknown.

This approach lets a family discuss America's startup economy without turning every technology headline into an investment thesis. You can learn about a product, examine a business and remain unconvinced about its investment terms. Those are compatible conclusions, not signs that the conversation failed.