Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteiTechGuides is reader-supported.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
A software company’s AI strategy is more convincing when it connects a named product to a real customer workflow, shows that customers pay for and keep using it, and explains how the economics can work as usage grows. Announcements, pilots, and large investment figures may be relevant, but none alone proves that AI is producing durable value.
Use the company’s annual report and other filings to test six things: product fit, paid adoption, returns and costs, competitive position, execution, and risk management. Treat management’s descriptions as claims to verify against reported results over time—not as independent proof that AI caused revenue growth or improved margins.
Start with the filings, not the announcement
For a public software company, begin with its latest annual report (Form 10-K for a U.S. domestic issuer) and compare it with earlier reports and subsequent quarterly filings. In the annual report, look for the business description, risk factors, and management’s discussion and analysis (MD&A). Search for terms such as “AI,” “artificial intelligence,” “machine learning,” “generative,” “model,” “inference,” and named AI products—but read the surrounding passages. A mention in a risk factor or strategy paragraph is not evidence of paid adoption.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Build a short record for each claim: what product or workflow is involved, what stage it has reached, what customer or financial measure supports the claim, and what costs or dependencies accompany it. Distinguish reported results from management expectations. A filing reflects what the company reports under applicable disclosure rules; it does not independently test product quality or prove that AI caused a financial result.
#1 Best Overall
1. Does the product solve a specific customer problem?
Connect the feature to a workflow
Identify the product capability and the task it is meant to improve. “AI across the platform” is less informative than a description of who uses a capability, what work it changes, and why that work is costly, frequent, or important. Look for evidence of a customer outcome—such as reduced manual effort or a faster process—and check whether the company quantifies it or provides only an illustrative example.
Check how mature the feature is
Separate generally available features from previews, limited releases, pilots, and future plans. A demo shows that a capability can be presented; it does not establish that customers can deploy it reliably, that it fits their workflow, or that they will pay for it. Look for statements about availability, deployment requirements, supported users, and customer adoption, and note when those details are absent.
2. Is adoption becoming paid, recurring use?
Trace the path from pilot to renewal
Follow the sequence from trial or pilot to initial deployment, production use, paid expansion, and renewal. Useful evidence includes customer counts, paid seats, usage, contract expansion, retention, and revenue—but only when the company defines what each measure counts. A deployment agreement or a customer relationship is not automatically a paying production customer. Check whether the company explains conversion rates, time to production, or implementation effort.
Recommended Free Tools
Read the revenue model carefully
Determine whether AI is included in an existing subscription, sold in a higher tier or as an add-on, billed by consumption, or supported by services. Bundling may encourage adoption, but it can make AI revenue difficult to isolate. Consumption charges may link revenue to use, while also making costs and customer bills usage-sensitive. If AI-specific revenue is not separately reported, say that it is not quantified in the disclosure; do not estimate it from broad cloud or subscription growth.
C3.ai, Inc.’s Form 10-K for the fiscal year ended April 30, 2026, reports the following company-specific measures. They illustrate what to track, not a sector benchmark:
| Fiscal year | Initial production deployment agreements | Subscriptions as share of total revenue | Professional services as share of total revenue |
|---|---|---|---|
| 2024 | 123 | 90% | 10% |
| 2025 | 174 | 84% | 16% |
| 2026 | 71 | 91% | 9% |
The company describes a shift toward engagements it considers more likely to produce targeted customer economic value and convert to production. It also describes usage charges based on virtual CPU/GPU hours after initial deployments. The agreement counts are not a direct count of AI product customers and do not, by themselves, show paid conversion, revenue, or customer value. The revenue mix is for the company overall; it does not isolate AI-specific revenue. Read the company’s definitions and explanations in the filing before drawing conclusions from changes between years.
3. Can the economics work as usage scales?
Compare value with the full cost of delivery
Look for evidence connecting AI use to revenue, retention, expansion, or customer value, then compare it with the costs of building and operating the capability. Relevant costs can include research and development, model licensing, compute, cloud hosting, data acquisition, implementation, and customer support. Ask whether the company discusses gross margins, cost trends, or the resources needed to serve increasing use.
Microsoft’s 2025 annual report says research and development expense rose $3.0 billion, or 10%, in fiscal 2025. The report attributes the increase to investments in cloud and AI engineering as well as Gaming and acquisition effects, and identifies AI training and other infrastructure costs within R&D. This is not an AI-only spending figure. A rising investment figure can show that the company is committing resources; without corresponding evidence of returns, it does not establish that the investment is productive.
Look for a credible path to better or sustainable margins
Ask whether the company explains how inference and infrastructure costs change with usage, whether pricing can cover those costs, and whether management reports improving or pressured margins as adoption grows. Pay attention to whether costs sit in R&D, cost of revenue, capital expenditures, or more than one category; accounting presentation can complicate comparisons. If the company does not disclose AI-specific costs or margins, treat that as an uncertainty rather than filling the gap with an estimate.
4. What does the company control—and depend on?
Assess the sources of advantage
Examine whether the company has distribution to existing customers, integration into important workflows, customer relationships, relevant data rights, a developer ecosystem, and security or compliance capabilities. These can help a product reach users or fit into established work. A claim of proprietary data is meaningful only if the company explains its rights to use that data and how it contributes to the product.
Map external dependencies
Identify which third parties supply essential models, cloud infrastructure, data, or distribution. Consider how a change in price, availability, service terms, or access could affect the product and its margins. Partnerships can provide valuable capabilities, but they can also shape bargaining power and economics.
Microsoft describes its OpenAI relationship as a strategic partnership and reports reciprocal revenue-sharing arrangements and rights relating to intellectual property and infrastructure in its 2025 annual report. Those are disclosed terms, not proof that the relationship will create a lasting competitive advantage. The investor’s question is what the arrangement enables, what obligations it creates, and how exposed the business would be if its terms or access changed.
5. Is the company executing repeatedly?
Shipping a feature is only one part of execution. Look for evidence that the company can deploy, maintain, and support AI products; convert experiments into paid production use; and do so without unusually high or unpredictable implementation effort. A repeatable process is more persuasive than a few showcase deployments if the business depends on broad adoption.
Compare claims across multiple quarters and annual filings. Check whether product availability, customer use, conversion, revenue, and cost disclosures progress in a consistent direction. Note when management changes definitions or emphasizes a new measure; a metric is harder to compare over time if its basis changes.
The SEC Investor Advisory Committee described AI as a strategic operational and competitive tool and highlighted integration challenges. Its recommendation, approved at the committee’s December 4, 2025, meeting, said that “the disclosures currently remain uneven.” That is a committee recommendation, not an SEC Commission rule or a company-specific finding. It is a reason to read company disclosures closely, not a substitute for evaluating execution.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →6. Are risks and controls described candidly?
Read AI-specific disclosures alongside ordinary risk factors. Check what the company says about data confidentiality and training practices, output accuracy, human oversight, cybersecurity, intellectual-property claims, regulation, workforce effects, and the possibility that customers build their own systems or replace the product. Look for both the stated controls and the risks that remain; polished descriptions of safeguards do not mean the risks have disappeared.
Best Value
Veritone’s 2025 Form 10-K says internal AI may improve productivity but cautions that “such productivity gains are not guaranteed.” It also identifies risks of exposing sensitive data and producing inaccurate or unreliable output. The example illustrates why productivity claims should be read together with the company’s qualifications and controls.
Legal requirements vary by jurisdiction and use case and can change. The SEC Investor Advisory Committee’s recommendation discussed an evolving regulatory environment; it is not a current legal checklist. For a company operating across markets, consider which jurisdictions and applications matter and verify applicable rules using current, authoritative sources before treating a regulatory statement as complete.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Compare companies on the same evidence
When comparing software companies, use common questions but account for differences in business model, reporting periods, and accounting treatment. A company selling consumption-based services will not necessarily report adoption or revenue in the same way as one bundling AI into a subscription.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
| Comparison axis | Evidence to examine | Question to ask |
|---|---|---|
| Product maturity and workflow | Named products, availability, deployment stage, workflow importance, and customer outcomes | Is the capability in production and useful for a meaningful customer task? |
| Paid adoption | Paid use, pilot-to-production conversion, renewal, retention, and expansion | Does adoption deepen and persist, or stop at trials and announcements? |
| Revenue model and disclosure | Bundling, tiering, add-ons, consumption fees, services, and AI-specific revenue disclosure | Can the company show how AI use becomes revenue, and what remains unquantified? |
| Costs and infrastructure | Development, inference, hosting, model, data, implementation, and support costs | Does the company explain how the economics change as usage grows? |
| Control and dependency | Access to customers, distribution, data, models, cloud infrastructure, and partners | Which important capabilities does the company control, and which depend on others? |
| Reliability and governance | Security, privacy, output quality, oversight, intellectual property, and regulatory risks | Are safeguards and remaining risks both described? |
| Evidence over time | Consistent measures and outcomes across filings and reporting periods | Do operating results support the claims, using comparable definitions? |
Less disclosure makes a comparison less certain and may signal lower transparency; the gap alone does not prove that a company’s AI business is weaker. Keep “not disclosed” separate from “zero,” and do not rank companies using figures that measure different things.
A repeatable review checklist
- Name the product and workflow. Record the AI capability, intended user, customer problem, availability stage, and evidence of customer outcomes.
- Trace adoption toward durable use. Look for defined evidence of production deployment, paid usage, conversion, renewal, and expansion; note what the company does not report.
- Identify how revenue is earned. Determine whether AI is bundled, tiered, sold as an add-on, billed by usage, or accompanied by services. Do not infer AI revenue from a broader category.
- Pair growth measures with costs. Track disclosed development and operating costs, infrastructure needs, margins, and cash generation. Mark AI-specific figures separately from company-wide totals.
- Map control points and dependencies. Note relevant customer access, data rights, distribution, model and cloud providers, and partnership terms.
- Test the story across time. Compare definitions and results in successive filings, and check whether earlier expectations are followed by reported outcomes.
- Read risks beside the claims. Record the company’s controls for privacy, security, reliability, intellectual property, and oversight, along with risks it says remain.
The SEC Investor Advisory Committee framed the challenge this way: “AI-related information can be material and of interest to investors, but the issue is how to sort the relevant information into operational categories that help inform investment decisions.” Organizing disclosures around products, paid use, economics, dependencies, execution, and risk makes it easier to distinguish evidence from promotion—without treating any single metric as an investment conclusion.
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.
