A financial research platform is a search engine over company filings, broker research, news and transcripts of earnings calls, in which a language model now reads the documents and answers questions directly. Investment teams, strategy teams and consultants use it. Filings are the reports that listed companies must submit to regulators and stock exchanges. In India, listed companies file them with the stock exchanges, under SEBI's rules. An earnings call is the call in which a company's management explains its quarterly results to investors and answers their questions.
The problem it solves is the amount of reading. An analyst who follows 30 companies may need to read hundreds of documents each quarter: results, annual reports, investor presentations, call transcripts and reports from broking firms. An ordinary keyword search finds the word "margin" thousands of times. It does not tell her what each company actually said about its margins.
So the platform does three things. First, it collects documents from many sources every day, converts them to text, and tags each one with the company, the date and the type of document. Second, it searches by meaning, not only by exact words. So a search for "price increase" also finds "we took a 3% hike". Third, a large language model reads the passages it found and writes an answer, with a link from each sentence to its source. This method, finding the passages first and then asking a model to answer only from them, is called retrieval-augmented generation (RAG). The links matter, because the analyst must check every figure before she uses it.
Take, for example, an equity analyst at a mutual fund in Mumbai who is studying four listed paint companies before the results season. She asks: "What did each company say about raw material costs and price increases in its last two calls?" Within a minute, the platform shows a table. One company expects its costs to fall by 2% next quarter. Two have raised their prices. One did not mention costs at all. Each line links to the paragraph of the transcript it came from. Without the platform, she would have read eight transcripts of about 25 pages each.
Many platforms also hold transcripts of expert network calls, which are interviews with former employees of companies. Some add alternative data or market research data.
How is it different from a financial data terminal? A terminal is built around numbers: live prices, financial statements and ratios. A research platform is built around documents, and what people said in them. Many analysts use both.