Most of the work isn’t the work
Where an analyst’s day really goes, and the part we built Analyst One to take back.
Ask anyone who’s had to do research or due diligence where their day went and you get a pause, then a list. Reading. Retyping. Hunting for the one line they know they saw somewhere. Almost none of it is the analysis. That bit sits on top, at the end, in the hour everyone actually cares about. There’s a lot of handle on the spear, and not much point.
We built Analyst One because that ratio is wrong. The work isn’t expensive to do, it’s just slow. Here is what that looks like, with rough times. They are typical of mid-cap PE, hedge fund and M&A work, and they move around a lot by deal.
Start with the file. It lands at 7pm and the partner wants a view by lunch. So you read it. All 150 pages, because the line that changes your mind is never on page 12, it’s in a footnote next to a table everyone glosses over. Call it 5 hours, maybe more, then a summary typed by hand, then an associate reads it again before it goes up. By the time anyone higher up sees a page, the team has spent most of a day on comprehension. Not judgement, not analysis.
Then the tables. And the charts, the graphs, the renderings, all of it. The model needs the numbers and the numbers are stuck in a PDF. So you retype them, or you copy-paste and watch it break on every merged cell and footnote. Half an hour a table, longer for the awkward ones, and one file can hold a dozen of them. Every small error rides quietly into the model, and the model is the thing the partner trusts.
Then the search. Not Ctrl-F on one file. It’s many files, a few tabs, an Excel model, maybe even an AI chat, plus your memory of who was on the last deal and what they flagged. The answer is in there somewhere, and finding it is its own afternoon. Worse is sourcing it. One claim in an IC paper means going back through the pile for the exact line, so under deal pressure the check gets skipped and the number goes in unproven.
And then you write it up. From scratch, restyled into the firm’s template by hand, and when a reviewer asks for a source the junior goes hunting all over again. Three rounds of edits, each one redoing the same formatting that was fine the first time.
None of this is exotic. It is the day. And it is the part we wanted to give back.
So here is what Analyst One actually changes. It gives you a first pass in minutes: a structured contents, the entities, the financials, the management quotes, each one with a citation you click straight back to. The tables come out with their structure intact and every cell linked to its page. You can ask the whole library a question by idea, not by filename, and get answers from every document with the source attached. When someone asks where a number came from, that is a click, not an afternoon. And the write-up starts in the firm’s format with the citations already on it, so the edits are about what it says, not how it looks.
The thing we keep coming back to is the citation. We can’t promise a model that never gets anything wrong. Nobody can, and I’d be wary of anyone who tells you otherwise. What we can do is make every answer checkable at the source, so the trust sits with the document and not with us. In this work, that is the whole game.
If you want the investor version of all this, it is one line. Analyst time is the most expensive input a firm has, and most of it goes on fetching and checking rather than deciding. Move even half of that off the analyst and you are not saving on stationery. You are buying back the hours your best people should be spending on the actual call. The logic holds whether the cheque is £50m or £500m.
We built Analyst One out of our own frustration with exactly this. If you have ever lost an evening to page 223, you already get it.
See it for yourself: felixresearch.com
