A first-look note for the sell-side desk. A weekly coverage update for the PM. A client-meeting brief for the wealth advisor. Configure the format once and the agent produces it in that shape on every release, with citations and white-label branding intact.
The problem with generic AI output
Research teams publish in specific formats. A buy-side firm has an internal note structure PMs expect. A sell-side desk has a first-look template compliance has approved. A wealth advisor has an email format clients are used to. The content changes every cycle. The format does not.
Generic AI produces generic output. A summary, in whatever shape the model felt like that day, with no relationship to the firm's actual deliverable. The analyst still has to translate it into the format the firm publishes in, which is exactly the work the tool was supposed to remove. By the time the translation is done, opening a blank page would have been faster.
Tailored Outputs solves the format problem. A Deep Research Agent is configured once with the firm's note structure, voice, and branding. From then on, every release produces a draft in that exact format, with citations preserved and the firm's logo, colour, and template in place.
What generic AI tools miss
The output is the deliverable. A research note is not a private document the analyst writes for themselves. It is a branded artefact that goes to a PM, a CIO, a sales force, or a client. The format carries firm reputation. Off-template output, however accurate the underlying analysis, isn't usable.
Most AI research tools stop at "here is the analysis". The user is then left to reformat into something publishable. Tailored Outputs goes the rest of the way: the analyst opens a draft that already looks like the firm's published work, with the right sections in the right order, the right tone, and the right styling. The judgment goes on top of a deliverable, not into a translation step.
The four-stage workflow
Stage 1: Configure
The format description is given in natural language: section headers, data presentation (tables versus prose), citation style, output length, tone. The firm's template, logo, and colour are uploaded once. Most users get this dialled in within a single sitting.
A custom style guide can be attached to capture firm-specific voice (the cadence of a sell-side desk's first-look note reads differently from a wealth advisor's client brief). The agent honours the style guide on every subsequent run.
Stage 2: Populate
When the trigger fires (a new earnings release lands, a weekly schedule ticks, an analyst runs the agent on demand), the agent pulls the relevant primary content and produces a draft in the configured format. Numbers slot into tables. Headlines slot into headers. Quote-level citations sit on every claim. Charts render in the firm's chart style.
This is the stage that compresses time. The reading, extraction, and templating happen in minutes instead of hours. The draft is ready in the firm's house format, not in generic AI prose that needs translation.
Stage 3: Polish
The analyst opens the draft and adds judgment. The thesis, the rating call, the framing the model can't generate, the line at the top of the note that explains why this quarter mattered. Most users edit rather than publish directly. That is the right decomposition: extraction and formatting are mechanical, narrative and judgment are not.
For follow-up questions that emerge during the polish stage, AI Analyst Chat answers against the same primary sources, with citations that flow back into the draft. Did management address the China question more directly than last quarter. Is this margin guidance consistent with the capital markets day commitment. The answers slot into the note rather than getting lost in a separate window.
Stage 4: Deliver
Outputs export as PDF, DOCX, Markdown, or plain text with citations and branding intact. Charts and tables export as images or structured data. Most teams flow the export into the firm's existing publishing system (Bloomberg NLRT, proprietary CMS, internal wiki) and finalise there. Some paste directly into client emails or PM Slack channels.
The branded artefact lands looking like the firm's existing work. That is the point.
Start your free evaluation
Analyze 15 leading companies immediately. No registration, no credit card, no sales call.
Common output patterns
The same configuration mechanism produces a wide range of recurring deliverables. The patterns most teams set up first:
- First-look earnings note (sell-side). Numbers vs. consensus, guidance delta, management tone, drivers to watch. Published in the firm's note template within minutes of the release.
- Weekly coverage update (buy-side). A standing template that summarises what moved on each covered name over the week, with source-linked callouts on anything material. Generated on a schedule or on-demand before PM meetings.
- Client meeting brief (wealth advisory). A per-client format covering the names the client holds, recent material changes, and the specific passages worth having ready if the client asks.
- Company primer for initiations. A standardised primer covering business model, segment economics, guidance track record, and management commentary, grounded in the company's full filing history. Shortens the initiation absorption phase.
- KPI tracking dashboard. A structured table that extracts specific operational metrics (DAUs, ARPU, customers, backlog, whatever the analyst tracks) from each new earnings release and updates the time series.
- Peer comparison update. A recurring cross-company comparison that pulls the same metrics or commentary points across a peer group and refreshes on each earnings release.
What compresses, and what doesn't
The compression is uneven across note types. Highly templated outputs (first-look earnings notes, KPI dashboards, peer comparisons) compress most, because the structure and the extractable content line up cleanly. Narrative-heavy outputs (thesis notes, industry commentary) compress less, because the value is in the narrative, not the extraction.
Tone can be specified in the agent configuration, and a custom style guide tightens the match further, but the agent produces a first draft, not a finished deliverable. Most analysts edit for voice and add judgment on final. The compression comes from removing the mechanical phase, not from replacing the analyst's writing. The underlying research on routine-research time reduction is here.
Where this sits in the wider workflow
Tailored Outputs is the output layer that sits on top of the rest of the platform. The earnings season workflow feeds it: a configured agent draft is what lands automatically when a new release is ingested. The management quality scorecard can be exported in the same templated formats. The shape of the deliverable is what changes across roles: the sell-side uses it for first-look notes against compliance review, the buy-side uses it for internal coverage briefs, the wealth advisor uses it for client-ready meeting briefs. The same configuration mechanism produces all three.
For background on how analyst communication formats differ across teams and channels, see the research repackaging post.
