# How Hedge Funds Consume Research in 2026

> An inside look at the morning of a buy-side analyst in 2026: the channels, the tools, the workflow gaps, and where AI-native platforms like [Loupe](https://louperesearch.com) fit in.

## The morning ritual

The buy-side analyst's morning in 2026 still starts the same way it did in 2016, with one difference: there is more information, from more channels, with the same fixed number of hours before the US open.

A typical morning involves five to seven surfaces:

1. **Email** — the inbound from 8 to 20 sell-side desks, plus independent newsletters and expert-network reminders.
2. **Broker portals** — the bank web apps where the PDFs actually live (Morgan Stanley Matrix, JPMorgan Markets, Goldman Marquee, etc.).
3. **Bloomberg Terminal** — for prices, news, and the chat function.
4. **AlphaSense or a competitor** — for searching transcripts and public filings.
5. **An internal notes system** — Notion, Coda, OneNote, or a custom workflow tool.
6. **A general-purpose AI chat** — ChatGPT, Claude, or Perplexity, used to fill the synthesis gap nothing else covers.
7. **Slack or Symphony** — for the firm's internal chatter and any sell-side IM relationships.

Each surface holds a piece of the picture. None of them holds all of it. The analyst is the integration layer.

## Channel 1: Email

Email is still the canonical inbound channel for sell-side research. A flagged broker email is still the moment of truth for "did the Street change its view on this name."

The problem is volume. A mid-sized hedge fund analyst covering software receives 200–400 broker-tagged emails before 9 a.m. Eastern. The triage step — open, scan, decide whether to read fully, decide whether to forward to the PM — is itself a 45-minute job. Most of it is throwaway. The signal-to-noise ratio is, by analyst self-report, somewhere between 5% and 15%.

Email was never designed to be a research database. There is no tagging, no entity recognition, no de-duplication when three banks publish on the same earnings print. Search across an Outlook archive is essentially useless for cross-document synthesis.

## Channel 2: Broker portals

When the email is a teaser, the actual research lives behind the bank's portal. Each bank's portal is a different UI, a different login, a different PDF viewer, sometimes a different MFA flow.

In 2026 most of these portals have a "summary" or "chat" feature bolted on. They are usable for a single document. They are useless for the cross-document question the analyst actually has: "what does the Street collectively think about this catalyst?"

## Channel 3: Bloomberg Terminal

The Terminal remains dominant for data, news, and chat. Its position is unassailable as the data layer. Its position as a written-research workspace is much weaker.

The 2026 Terminal includes an LLM chat surface. In practice analysts use it for narrow, well-defined queries: "summarise this 8-K," "compare these three companies on EBITDA margins." For the broader workflow — read every broker note on a name, synthesise the consensus view, surface the dissenters — the Terminal is not where the work gets done.

## Channel 4: AlphaSense and search platforms

AlphaSense (and competitors like Sentieo and Daloopa) own the use case of searching public filings, transcripts, and expert-call libraries. They are the de facto answer to "find me every mention of X in earnings calls in the last two quarters."

What they do less well: bilateral broker research from the full Street, feed-driven workflows organised around watchlists and portfolios, and a publishing surface for the analyst's own work.

## Channel 5: Internal notes

Every fund has a system. Most of them are Notion, Coda, OneNote, a wiki, or a Frankenstein of Google Docs. The point of the system is to give the PM and the analyst a shared, searchable memory: "what did we conclude about this name three quarters ago?"

The notes system is rarely connected to the research feed. The analyst manually copies the salient bullet from the broker note into the internal doc. The link is preserved if the analyst is disciplined. Often it is not.

## Channel 6: General-purpose AI chat

In 2026 essentially every buy-side analyst uses a general-purpose AI chat in the workflow. ChatGPT, Claude, and Perplexity are the leaders.

What they are used for, in order of frequency:

1. Summarising a single document the analyst has pasted in.
2. Drafting the language for an internal memo.
3. Answering "explain this concept" questions.
4. Cross-checking a number or an entity name.

What they cannot do: see the firm's licensed broker research, see the analyst's private notes, see the portfolio's positions. The most relevant content for institutional research is not on the open web. That is a structural ceiling for general-purpose chat.

## Channel 7: Chat and messaging

Symphony for institutional IM with the sell-side, Slack for internal. The interesting research often arrives here first: a sell-side trader pinging a colour read, an internal analyst flagging an alternative-data signal, a Slack channel summarising the day.

The chat surface is high-signal and ephemeral. It rarely gets archived into the research workflow.

## The integration problem

The seven surfaces above produce a 2026 workflow that looks roughly like this:

- 30 minutes triaging email.
- 45 minutes reading PDFs in broker portals.
- 20 minutes on Bloomberg for news and data.
- 15 minutes on AlphaSense for one or two targeted searches.
- 10 minutes updating internal notes.
- 15 minutes back-and-forth with a general-purpose AI for summarisation and synthesis.
- Ambient chat throughout.

Total morning routine: roughly two hours before the actual investment work — building a model, talking to a PM, sizing a trade — begins. For a fund running 30 names per analyst, this is the gating constraint on coverage depth.

## Where AI-native research platforms fit in

The bet a platform like [Loupe](https://louperesearch.com) is making is that the integration layer should not be a human. It should be a workspace built around three things:

1. **A licensed corpus** — bilateral agreements with banks and independent publishers so the AI can see the research the fund pays for, not just the open web.
2. **A feed-driven surface** — push, not pull, organised around the watchlist and the portfolio, so the analyst sees what matters before asking.
3. **A grounded chat surface** — an AI assistant that answers from public filings, licensed broker research, and the analyst's own private context, with citations back to the source.

This is the three-layer context model: public + licensed + private. It is what a generalist chat is structurally locked out of, and it is what a data terminal was never designed to provide.

## What this means for the next two years

Three predictions for buy-side research consumption through 2027:

1. **Email volume goes down, not up.** As more banks make their research available through aggregators with bilateral agreements, the daily email blast becomes a notification, not the canonical channel.
2. **The PDF starts to die as the unit of research.** Structured chunks — earnings preview, rating change, model update — replace the 40-page report as the atomic unit that flows through the workflow.
3. **The "morning prep" surface consolidates.** Today it is seven surfaces and a human integrator. By 2027 the leading hedge funds will run morning prep through one or two surfaces, with AI doing the cross-document synthesis that the analyst does manually today.

The funds that win that consolidation buy back the most scarce resource on the buy-side: analyst hours. Two hours of morning prep recovered per analyst per day, across a 30-person research team, is roughly 60 hours a day — enough to cover ten more names with the same headcount, or to spend ten more hours per name on the ones already in the book.

That is the real prize. The platform that delivers it owns the next decade of institutional research.

---

*[Loupe](https://louperesearch.com) is the AI research workspace for high finance. Currently invite-only — join the waitlist at [louperesearch.com](https://louperesearch.com).*
