AI for equity research is software in which an AI model reads company filings, earnings call transcripts and reports, drafts research notes, and answers an analyst's questions from that data. Equity research means studying companies to decide whether to buy, hold or sell their shares. The main users are on the buy side, which means the firms that invest money, such as mutual funds, hedge funds and insurers.
Why is it useful? An analyst at a fund may follow thirty companies. Every quarter, each company publishes its results. Most also hold an earnings call, a phone call in which the company's managers explain the results and answer analysts' questions. The transcript of one call can run to thirty pages. In a results season of about six weeks, the analyst must read all of this, update a spreadsheet model for each company and write a note for the fund manager.
Say an analyst covers six paint companies. She asks the software: "Which of these companies said that raw material prices reduced their profit this quarter, and by how much?" The model reads the six transcripts and the six sets of results. It answers in a table, and each figure is linked to the page and line that it came from. She opens two of the links to check them. Then she asks the model to draft a one-page note on the company that did best.
The link to the source is the most important feature. A language model can produce figures that look right but are wrong. So the analyst checks the draft against the sources before the note goes to the fund manager. The model drafts, and the person approves. This pattern is called human in the loop.
The model needs reliable data. It reads company accounts and analysts' estimates from a financial data terminal or a financial research platform. The analyst's final note goes to the fund manager, who may then place an order in the fund's investment management system.
For many years, much of this reading and data entry was done by large teams of people in India. AI models now do more of the reading, and the large data companies are building such assistants into their own products.