Most of what gets called equity research is not analysis. It is retrieval. Finding the segment disclosure, pulling the guidance language out of a transcript, checking whether the risk factors moved since last year, tracking a metric back ten quarters. The judgment part, the part you are actually paid for, usually takes a fraction of the time that the fetching does.
That ratio is what we built Hudson Labs to fix. This is the workflow we walk new users through, written out in one place. It works whether you cover twelve names or two hundred.
Start with one company
Company research begins with a ticker. Pull one up and you get the snapshot on a single page: the stock chart, key financial stats, and direct links to the 10-K, 10-Q, earnings release, transcript, presentation and DEF 14A. Filings, transcripts, financial metrics, red flags and peer comps all sit together, so you are not opening six EDGAR tabs to answer one question.
From there you ask the Co-Analyst whatever you need and get a sourced answer back. Four queries come up constantly:
The earnings snapshot. The release and the call in one view: what beat, what missed, what management flagged. Example output
Every change in guidance. Each shift in forward guidance, sourced to the exact line. Current guidance and long-term guidance
An operating metric over ten quarters. Same-store sales and traffic, quarter by quarter, ready to transpose into a model. Example output
Accounting changes. Where a company changed how it accounts for something, with the language quoted. Example output
Two things speed this up. Every tear sheet ships with suggested prompts built for that specific company, and any query you write can be saved as a template and reused across your whole coverage list. If you want to see what a finished piece of work looks like, here is a complete project on CMG.
One note on phrasing. The query box responds better to search terms than to full questions. "Artificial intelligence" will get you further than "what was the impact of AI demand on Broadcom last year and this year." Think of it as a topic you are pointing the system at rather than a chatbot you are talking to.
Take the same question across the universe
Company research handles one ticker. Market Intelligence runs the same kind of question across roughly 10,000 companies, which is where most thesis work actually starts.
The steps are the same every time:
- Ask the question in plain English. Something like "which companies are raising prices in response to tariffs."
- Choose your sources and period. Earnings calls, conference calls, releases, 8-Ks, 10-Ks, 10-Qs, over whatever window you care about.
- Narrow it. Layer on metric filters like market cap, sector, gross margin or revenue growth, or run it against a watchlist to bring 10,000 plus companies down to the ones inside your mandate.
- Run it, then sort, export or save. Saving matters more than it sounds. Next quarter you re-run the screen instead of rebuilding it.
Screens that tend to earn their keep: companies growing gross margin three straight quarters, consumer demand commentary broken out by income cohort, names flagging longer lead times or supply constraints, and year over year backlog increases above 50 percent.
Every result links back to its source, so you can defend the screen itself and not just the individual answers inside it.
Don't start from a blank prompt
A blank prompt box is the slowest part of any AI research tool. Our research library exists to skip it. There are 90 plus curated queries and screens built by our team, organised by category: deep dive, industry landscape, market and macro, risk and red flags, AI and emerging technology, accounting, and more.
A few that get cloned often:
- The AI ecosystem stack market map, covering 500 plus public and private companies with the bottlenecks flagged at each layer
- A GLP-1 market map with milestones, phase and performance detail
- Private credit redemptions, performance and tone, broken out by type
- Tech and industrials reporting backlog or RPO increases above 50 percent versus prior year
The method is the same in each case. Find something close to what you need, clone it into your workspace, then edit the prompt or the filters to fit your ticker or your thesis. The cloned version becomes a template the rest of your team can reuse.
Put the recurring work on a schedule
Some research repeats on a calendar. Earnings summaries every quarter, risk redlines when a new annual filing drops, a thematic screen every month. That work does not need a person triggering it.
Agents run it on a schedule and email you the output. The preconfigured ones you can turn on in a click:
- Earnings summary. When a new call is published, run a consensus beat and miss summary and email the result.
- 10-K and 10-Q key risks. When a new annual or quarterly filing lands, redline the risk section against the prior filing.
- Price drop over 5 percent intraday. When a covered name drops more than 5 percent in a day, run two explainer queries, one over recent filings and one over web news.
- Consumer demand trends. A monthly thematic screen across earnings and conference calls.
You can also build your own from the topic monitoring, document summary, earnings preview or price monitoring builders. Set what to watch and how often, and it runs in the background. Everything you have built lives in one place.
The part that makes it usable in finance
None of the above is worth much if you cannot show your work. Every answer on the platform traces back to the exact filing and passage it came from. Select any part of an answer and click "cite this text" to jump straight to the source. If you want to check it yourself, keyword search across the underlying filings and transcripts in the sources panel.
An analyst cannot put an unverifiable number in a memo, and compliance will not let them try. Sourcing is what allows the output to leave your desk at all, which is why it sits underneath everything on the platform rather than off to the side.
What this looks like in practice
An analyst on the platform needed same-store sales, segment level margins and unit economics out of a retailer's 10-Q, fully sourced and citeable. Line by line, that is a few hours of work. It took 20 minutes.
"I use Hudson Labs to take what was a two or three hour process down to a 20-minute process." Brett Caughran, Founder, Fundamental Edge
Hudson Labs is used by analysts at Fortune 100 companies, top five law firms, and investment managers overseeing more than $600B in assets. Our work has been covered by Bloomberg, the WSJ, Forbes, CNBC, the Financial Times and Institutional Investor.
Try it on your own coverage
The fastest way to judge any of this is to run it against a name you already know well, where you can check the output against what is in your head.
Start a free trial or book a 30-minute walkthrough if you would rather have someone show you the workflow on your own names.
