Up to about 40% of routine research time can be reclaimed by automating the mechanical reading phase. The interesting question is not whether that compression is real. It is what a team chooses to do with the recovered capacity.
The structural tension
Research budgets have not kept pace with coverage ambition. The CIO wants more names, more adjacent sub-sectors, more non-US exposure. The headcount line has not stretched to match. The traditional resolution is one of two unpleasant choices: stretch the analysts thinner or under-deliver on coverage. Either way, the team carries the cost of that compromise on its calendar.
The compromise softens when the mechanical phase contracts. It does not disappear. It softens, and what the team does with the softening is a leadership choice that deserves to be made explicitly.
The headline number, from Marvin's research: up to around 40% of routine research time. That covers the filings, earnings calls, and press releases every analyst processes regardless of conclusion. Not the thesis work, not the judgment, not the client conversations. The mechanical surface around them.
What generic AI tools miss
The lazy framing for any AI tool in this space is "do the same coverage with fewer people". Plausible-sounding to a CIO. Mostly wrong on the ground.
The compression is in the reading and extraction phase, not in the parts of the job that distinguish a good analyst from a mediocre one. Pointing a generic summariser at filings and asking it to replace headcount conflates the two. Marvin's claim is narrower and more honest: the mechanical layer compresses, and the time recovered is genuinely usable elsewhere. Where that time goes is a desk-level decision that depends on the firm's strategic priority, the shape of the existing coverage, and the binding constraint the desk is actually trying to relax. There is no single answer to that question, and any vendor offering one should be asked for the sample size.
Where the time actually comes from
Three places, in roughly the order they show up on a typical analyst's week.
Earnings coverage is the largest single contributor for most covering analysts. Twenty names, sixty primary documents, three weeks. The detailed shape sits in the earnings season solution.
Ongoing monitoring is the steady-state workload between cycles: ad-hoc disclosures, news, press releases, peer commentary, and tracking guidance against delivery. Guidance Tracking handles the promise-vs-delivery layer automatically, and the management quality assessment describes the structured workflow it feeds.
Initiation absorption is the multi-week reading required to be credible on a new name. Deep Research Agents produce a structured primer covering business model, segment economics, guidance track record, and management commentary. The absorption phase shortens. The thesis work and model build do not.
For five concrete workloads with time-savings breakdowns, see the automated workflows post.
Three ways teams redirect the time
More names per analyst
The most direct redirect, and the one most teams reach for first. How many more depends on how templated the existing coverage is and how adjacent the new names are. Expansion into thematically adjacent sub-sectors works better than expansion into unfamiliar domains, because analyst judgment still applies. The tool does not create domain expertise it doesn't have access to. A platform analyst who already knows software economics will absorb a new SaaS name faster than they will absorb a railroad.
Faster initiation ramps
For sell-side teams under pressure to initiate quickly when sectors shift, the initiation ramp is often the binding constraint. The absorption phase shortens when the primer is generated rather than assembled. The analyst still adds the thesis and builds the model. That part does not get faster. What gets faster is the calendar between the coverage decision and the published initiation.
Going deeper on names already covered
Not every team wants more names. Some want existing analysts to spend more time on the thesis calls they already cover. For buy-side teams where depth is the edge, this is often the right redirect. The detail sits on the buy-side use-case page, where the argument is that compressing the reading phase frees the calendar for the off-page work (management calls, channel checks, peer conversations) that actually generates alpha.
Start your free evaluation
Analyze 15 leading companies immediately. No registration, no credit card, no sales call.
What we know, and what we don't
What we know: the compression on mechanical research time is real, and it comes from a specific part of the workflow.
What we don't know: which of the three redirects above outperforms at the firm level, because we have not published comparative data across modes. Teams we talk to operate in each of them. That's the current state of the evidence.
What we specifically don't promise: a fixed number like "X to Y additional names per analyst". That number depends on the team, the sector, the template, the conviction threshold for new coverage, and a half-dozen other inputs we don't have visibility into. Any vendor who offers it as a headline number should be asked for the sample size.
How to actually evaluate the idea
The honest path is bottom-up. The case for bottom-up adoption over top-down rollout is detailed elsewhere, and the team implementation guide walks through the rollout. The short version:
- Pilot with one or two analysts on existing coverage. The tooling is single-player by default, so this does not require a firm-wide license or a procurement cycle.
- Track where the time actually gets saved across a full earnings cycle before drawing conclusions. Compression is uneven across workflows, and the per-name distribution is the input that should drive the redirect decision.
- Decide what to do with the recovered capacity based on the actual numbers, not the assumption. Expanding coverage is the most commonly assumed redirect. It isn't always the one with the biggest payoff.
Non-US expansion specifically
Coverage expansion often runs into the non-US question: the names the team wants to cover are in markets where the filing infrastructure is less structured than SEC EDGAR. Any company that communicates with investors in English is supported, regardless of where it is listed, with filings tracked, calls transcribed, and decks parsed to the same standard. The global company coverage overview walks through how non-US names are added, and Pro plan customers can request custom coverage for up to 50 uncovered companies.
