Independent research notes on AI and investing
UO Solutions examines what AI tools can and cannot do for individual investors who research US-listed companies. We publish our test methods and results in free articles. We also offer paid one-on-one consultations on research process — not on which stocks to buy.
What this site provides: educational articles and documented tests about AI-assisted stock research methods.
What it does not provide: investment advice, stock recommendations, price targets, or trade signals. We are not a registered investment adviser.
Cost: all articles are free. Consultations are paid and scoped in writing before any work begins.
Operated by: UO Solutions, LLC, 2875 NE 191st STE 801, Miami, FL 33180. Last updated: August 1, 2026.
Most individual investors do not have an information problem. Company filings, transcripts, and data are free and abundant. What they have is a process problem: too much text, too little time, and no reliable way to tell a useful summary from a confident mistake. AI tools stepped directly into that gap, and the marketing around them claims far more than the tools deliver.
In our own reading tests, the pattern repeats. A general-purpose chatbot will condense a 200-page annual report into ten pleasant paragraphs in seconds. It will also, on a regular basis, attribute a figure to the wrong fiscal year, drop a qualifying sentence that changes the meaning of a risk factor, or state with confidence something the filing does not say at all. None of these errors look like errors. That is what makes them expensive for an investor who stops checking.
Our judgment is that AI is genuinely useful in research, but only in specific positions of the workflow: locating sections, reformatting tables, drafting comparison questions, flagging where two disclosures disagree. It is much less reliable at the step that matters most, which is deciding what the numbers mean. That step still belongs to the reader. Everything on this site follows from that division of labor.
Our source hierarchy is fixed. First, primary documents: SEC filings retrieved from EDGAR, regulator bulletins, and official investor-protection material. Second, reproducible tests we run ourselves, with prompts and outputs kept on file. Third, academic work on language models and financial text. Company marketing pages are used only to document what a tool claims, never to support what it does.
When a tool's output conflicts with the filing, the filing wins, and we show both so you can see the gap. When two credible sources conflict with each other, we say so instead of picking the one that reads better. Every article carries a publication date and a review date, because tool capabilities and pricing change faster than our opinions do. Material we cannot verify does not get published; it gets dropped or marked as unresolved.
Articles are written and reviewed by Donald Ramirez, Editor, reachable at donald_ramirez117@gmail.com. Corrections go to the same address and are noted on the corrected page rather than silently edited.
The clearest sample of our working method is the four-tool test on a single 10-K, which includes the prompts used and the errors found. If you want to see how a claim on this site was produced, ask; the test files are kept.
One annual report, four popular tools, identical questions. Three of the four attached at least one figure to the wrong fiscal year, and all four stripped qualifiers out of the filing's carefully worded risk factors. The full method, prompts, and error log are included so you can repeat the test with your own tools.
Why the registration status of whoever answers your question changes what they are allowed to say, and what you should expect.
Data sources, update frequency, refund terms, and the one question about backtested results that vendors dislike most.
For locating information, mostly yes. For interpreting it, no, not without verification. In our tests the tools were consistently better at "where does this filing discuss lease obligations" than at "are lease obligations growing faster than revenue," because the second question requires judgment about which figures belong together. Our documented test shows the specific failure patterns.
No. That is investment advice, and providing it for compensation generally requires registration as an investment adviser, which we do not hold. You can verify any adviser's registration status through the free search tool at Investor.gov, run by the US Securities and Exchange Commission. What we can do is help you build a research process that relies less on anyone's tips, including ours.
You receive a reply within two business days. The first reply does not sell anything; it asks what you are trying to improve and states whether your request falls inside our scope. If it does, we send a written scope with deliverables and price before any commitment. If it does not — for example, if you are asking for stock picks, tax advice, or portfolio review — we say so and decline. About a third of inquiries end there, and that is fine.
No, and nothing on this site should be read as personalized investment advice. We are an educational research publisher and process consultant. The distinction matters legally and practically: we are not permitted to recommend specific securities for your situation, and we structure our services so we never need to.
Each engagement is quoted in writing before work starts. The quote depends on real variables: how many tools your process involves, whether you need a working session or a written brief, and how much documentation already exists. The services page explains these variables in detail. What we can commit to here: you will never be asked to pay before the scope and price are agreed in writing.
No. These are the areas where AI-generated marketing does the most damage and where a small team cannot test responsibly. Limiting scope is a deliberate editorial choice, not a gap we plan to fill.
