- The assistant answers questions such as what drove this quarter's profit from the articles, filings and clips your team already follows, and shows the passage behind each answer.
- Enterprise h2oGPTe indexes your sources in a private collection and writes grounded, cited answers inside your environment.
- Answers can come back as a table, ready to compare several companies or quarters in one view.
- An agent can turn a public economic data series into a dashboard with trend notes, for a person to review before it is shared.
Find the fact behind a market move
Your analysts, economists and communications teams follow dozens of sources every day: wire stories, earnings coverage, filings, broadcast clips and economic data releases. Each one takes time to read, and the fact that matters is often one line in a long piece.
A general chatbot is not the answer. It cannot show you which article a number came from, it can mix up companies or quarters, and your research questions leave your environment.
Most of what your team reads is unstructured text, not a tidy database, so the answer usually means opening several documents and comparing them by hand.
How financial news analysis works
The pattern is retrieval-augmented generation: an assistant finds the passages that answer a question in your own sources, then writes the answer from them and cites them.
Step 1: build the collection
Add the news articles, filings, transcripts and clips your team follows to a private collection in Enterprise h2oGPTe. Collections stay private until you share them with named users or groups.
Step 2: find the right passage
Semantic search finds matches by meaning as well as by exact company names and tickers, so a question about a subsidiary still finds the filing that names it.
Step 3: answer with the source
A large language model writes a grounded answer and shows the references it used, so an analyst can check the line before it goes into a briefing.
Step 4: shape the output
Ask for a table that compares companies, quarters or factors, or a short summary for a morning note.
Step 5: add the data series
An agent can load a public economic series, clean it and chart it, so the story and the numbers sit side by side.
Keep research private while it moves fast
Enterprise h2oGPTe runs in four deployment options: on-premises, air-gapped, private cloud and cloud, so your research questions and sources stay in the environment you choose.