Skip to main content
Chat is your AI research assistant inside AllMind. You ask a question in plain language, choose how deep you want it to go, and decide which sources it should draw on — and it plans the work, searches, reads, calculates, and writes back a grounded answer with citations you can open and check. It can pull live market data, filings, transcripts, research, news, the web, and your own documents; run real financial calculations; build charts and tables; and produce finished deliverables like PowerPoint decks, Excel models, Word documents, and multi-section research reports — all in a single conversation you can return to anytime. Find it in the sidebar under Chat — it sits at the top of the navigation, next to your Dashboard.

​
When to use it

Reach for Chat when:
  • You want a fast, well-sourced answer to a finance question — a price or metric, a definition, a quick read on a company, sector, or filing.
  • You want to work conversationally — ask, see the answer, then refine with follow-ups in the same thread.
  • You want the assistant grounded in your materials — your Data Room documents, an uploaded file, or a connected account — not just the open web.
  • You need a deliverable out of the conversation: a chart, a table, a model, a deck, or a report.
  • You want to stay in control while it works — watch its steps, redirect it mid-run, or stop it.
Use something else when:
  • You want to kick off long, autonomous research and check back later → Agent Studio. Chat and Agent Studio share the same underlying assistant; Chat is the live, conversational way to use it, and Agent Studio runs it hands-off on a task.
  • You want to find and gather documents or sources first → Document Search, then bring them into Chat or a Data Room.
  • You want the same set of questions answered consistently across many companies or documents → Grids.
  • You want a long, formatted, templated report as the primary deliverable → Reports (though Chat can also generate a sourced report and help you edit one — see How it works with other features).

​
How to use it

1

Start a conversation

Open Chat and type your question in plain language — or pick one of the suggested starter prompts. Be specific about the company, period, or angle you care about. A new conversation is created automatically and named for you based on your question, so it’s easy to find later.
2

Point it at the right sources (optional)

Open the sources controls to choose where the assistant should look — your Data Room (whole room or specific folders), the document library (by type, e.g. filings or transcripts), standardized financial data, the web, and a connected data warehouse if your organization has one. Narrow the sources for a focused, scoped answer; leave more on for broader coverage.
3

Add specific context (optional)

Type @ in the message box to mention a specific company (by ticker or name) and pin it to an exact fiscal year, quarter, and data type — fundamentals, a transcript, a filing, research, news, and more. Attach files to the message, or pull items in from a connected account. Each item you add is listed on your message, so you can see exactly what the assistant was given.
4

Choose how deep to go

Pick a mode — Instant, Balanced, Agent, or Deep Research — to trade speed for thoroughness (see below). Agent is the default and handles most analytical work.
5

Send, and watch it work

The assistant streams its progress live — the steps it’s taking, its thinking, a running plan or task checklist for bigger jobs, periodic checkpoint summaries, and the sources it’s pulling — then writes the answer with inline citations. While it’s working you can steer it (type a new instruction to redirect it without starting over) or stop it.
6

Review, follow up, and take the result with you

Open any citation to see the supporting material. Ask a follow-up in the same thread to go deeper or change direction — the conversation keeps its full context. Then copy the answer, export it to PDF or Word, email it, or download any file the assistant produced.

​
Choosing a mode (how thorough vs. how fast)

Chat offers four modes, trading speed for depth. The default is Agent.
  • Instant — fastest; for a single fact, a price, or a quick check.
  • Balanced — a quick but more capable pass; good for everyday questions and short summaries.
  • Agent (default) — the standard workhorse. Handles real analytical work: multi-company comparisons, earnings and filings research, valuation and modeling, charts and tables, and producing client-ready Excel, PowerPoint, and Word deliverables. Takes a few minutes.
  • Deep Research — the most thorough. It plans the work, runs several lines of research at once, and digs deepest — best for open-ended questions, sector and industry deep dives, peer studies across several companies, and investment theses. It takes the longest.
Rule of thumb: start with Agent for real analysis; step up to Deep Research when the question is broad or important enough to justify the wait; drop to Instant or Balanced for quick lookups and follow-ups.
Agent and Deep Research runs keep going if you close the tab — you can leave, switch devices, and return to the finished result. Instant and Balanced are meant to be quick and complete while you wait, so they don’t survive leaving the page mid-answer.
Chat’s “Deep Research” mode is its deepest setting within a live conversation. It is not the same as the standalone Deep Research product, and not the same as Agent Studio (which runs the same assistant autonomously on a task). People mix these up — the distinction is live-and-conversational (Chat) vs. autonomous-and-hands-off (Agent Studio).

​
Capabilities & key choices

Chat is a single assistant with a wide toolkit. The most useful things to know you can do: Ask, compare, and analyze. Look up a company, compare several, dig into an earnings call or a filing, screen a sector, build a valuation, or ask an open-ended research question. Modes (above) set how deep it goes. Let it follow a professional method automatically. The assistant draws on a large library of built-in analytical playbooks — valuation models (DCF, LBO, comparable companies, merger and three-statement models), due-diligence checklists, investment memos, earnings previews, sector studies, and many more — and applies the right one based on your question, so you get a structured, professional approach without having to spell out every step. Scope the sources. Decide exactly where the assistant searches:
  • Data Room — your saved documents; enable the whole room or pick specific folders.
  • Documents (library) — the curated document library, filterable by type: SEC filings, SEDAR (Canadian) filings, earnings-call transcripts and slides, investor-relations materials, press releases, news, sector data, ESG reports, and third-party research reports.
  • Standardized financial data — structured market, fundamental, estimate, and ratio data.
  • Web search — live web results for recent developments (on by default; turn off for a closed-universe answer grounded only in your chosen sources).
  • Connected data warehouse — if your organization has connected one, the assistant can query approved tables (and search an index of them across your chats); this appears only when it’s available to your organization.
Add precise context with @. Mention a company by ticker or name and pin it to an exact fiscal year and quarter and a data type (fundamentals, transcripts, slides, filings, news, press releases, research, sector data). A guided picker lets you build the mention from menus. Bring in your own material. Attach files to a message (drag-and-drop, file picker, or paste a screenshot), or import from a connected account: cloud document storage (personal and corporate) and email. For email you can set safe-list domains to control what’s searchable. Highlight text anywhere on the page and add it as context for your next question. Save and reuse instructions. Apply a Skill or a saved prompt template from the message box to steer the assistant for a question — your own, or one from AllMind’s built-in library — so you get consistent results without retyping. You can also build reusable prompt templates with fill-in fields (company, period, data type, and so on). Polish your prompt. Use the prompt-improve option to rewrite a rough question into a clearer, better-scoped one before sending, and open an expanded editor for long prompts. You can also dictate a question by voice. Get real calculations, not just prose. The assistant runs genuine quantitative work — DCF, LBO, NPV/IRR, comparable-company analysis, regressions, volatility and correlation, growth metrics (CAGR, year-over-year, margin bridges), and fixed-income math — so numbers are computed, not estimated. Get visuals. It can generate interactive charts inline (line, bar, area, pie, scatter, waterfall, radar, and candlestick, plus simple diagrams and flowcharts) and clean data tables; charts can be exported as an image. Produce finished documents. Ask for a deliverable and download it: PowerPoint decks, Excel models (with live formulas and standard banking formatting), Word documents, and multi-section research reports where every figure traces to a cited source. For presentations you can pick a template style from 25+ options — a premium AllMind house style (the default) plus styles modeled on major investment banks (Goldman Sachs, J.P. Morgan, Morgan Stanley, and more) and clean general styles. The assistant can also edit an existing PowerPoint, Word, or Excel file you provide. Generated files appear as a card you can preview online or download. Have it email a result. Ask the assistant to email a summary or a generated file to specific recipients; it shows a confirmation preview (recipients, subject, attachment) before sending. Stay in control of a run. While the assistant is working you can steer it — type a new instruction and it adjusts course while keeping the conversation so far — or stop it entirely. Take notes alongside the chat. A notepad panel sits beside the conversation: jot notes, add an answer to your notes, and let the assistant update your notes as the work progresses. Manage your conversations. Every chat is saved to your private history, newest first and grouped by recency. You can reopen any past conversation with its full transcript — answer, reasoning, and citations intact, with the step-by-step activity restored — search your history by keyword, start a new chat, or delete one. Deleting is immediate and permanent (there’s no trash), so export anything you want to keep first.

​
What it can access & produce

It can draw on:
  • Live market & company data — prices and history, financial statements (annual and quarterly), valuation multiples, a deep library of financial ratios, segment and geographic breakdowns, consensus estimates, price targets and ratings, technical indicators, peer and relative performance, seasonality, ETF holdings, index and commodity quotes, corporate events and earnings calendars, leadership and governance, ESG scores, dark-pool activity, post-earnings reactions, and market-sentiment signals.
  • Filings, transcripts & research — US and Canadian regulatory filings, earnings-call transcripts and presentation slides, broker and independent research, investor-relations materials, press releases, sustainability reports, and expert-interview transcripts (rolling out gradually to teams).
  • News, macro & specialty data — market and company news, a large library of macroeconomic series, prediction-market probabilities, M&A deal data, and supply-chain / customer-supplier relationships.
  • Alternative data — consumer-spend trends, foot traffic, app usage and engagement, and hiring and headcount signals that can lead reported financials (when enabled for your organization).
  • Your own materials — your Data Room documents, files you upload to the conversation (including images it can read), and connected accounts (cloud storage, email, a data warehouse).
  • The web — live search for recent news and developments.
It can produce:
  • A written answer with inline citations back to every source it used.
  • Interactive charts and data tables rendered in the conversation.
  • Downloadable files — PowerPoint decks, Excel models, and Word documents (and edits to ones you provide).
  • Multi-section research reports where every figure is tied to a verified source.
  • An emailed copy or an exported PDF / Word version of an answer.

​
Tips & best practices

  • Be specific. Name the company, the period, and the output you want (“…and build a 5-slide summary deck” or “…show it as a table”). Specific asks get sharper answers.
  • Scope your sources to the job. For a focused, defensible answer, narrow the sources to exactly what should count (e.g. just this company’s filings and your Data Room) and turn off the web. For broad discovery, leave more on.
  • Ground it in your own documents. Point it at a Data Room folder or attach files when you want the analysis based on your materials rather than the open universe.
  • Use @-mentions for precision. Pinning the exact ticker, period, and data type removes ambiguity and speeds things up.
  • Match the mode to the stakes. Use Agent for real analysis, Deep Research for big or high-stakes questions, Instant/Balanced for quick checks.
  • Refine with follow-ups instead of starting over — the thread keeps context, so each follow-up builds on what’s already there.
  • Steer instead of restarting. If a long run drifts, type a redirection to adjust it without losing the work so far.
  • Open the citations. The source trail is the point — verify the key numbers before you rely on them.

​
Limits & things to know

  • Deeper modes take longer — that’s expected; they’re doing more thorough, multi-step work. Agent and Deep Research also keep running if you close the tab and can be resumed; Instant and Balanced are meant to finish quickly while you wait.
  • File attachments: each file must be 12.5 MB or smaller, and supported types are PDF, Word, Excel, PowerPoint, CSV, plain text, common images (PNG, JPG/JPEG, WEBP), and MP3 audio. You can attach several files to a single question.
  • Web search is on by default. Turn it off when you want an answer grounded only in your chosen sources.
  • A connected data warehouse appears only if your organization has set it up, and the assistant can query only the tables your organization has approved.
  • Conversations hold a very large amount of context, so long threads and big documents are fine — but the fastest modes (Instant/Balanced) work from a smaller memory window, so for very large documents or long histories use Agent or Deep Research. Start a new conversation for a clean slate on a new topic.
  • Answers reflect what the assistant could access at the time it ran, given the sources you enabled, and connected-account and alternative-data sources only work once you’ve connected or enabled them.
  • Your chats are private to you. There’s no built-in way to share or hand off a conversation to a colleague; to share an answer, export it to PDF/Word or email it.
  • Chats stay in your history until you delete them (there’s no automatic expiry), and deleting is permanent. Download anything you need to keep long-term.
  • “Deep Research” mode here is the deepest setting inside a live chat — not the standalone Deep Research product, and not Agent Studio.

​
How it works with other features

Chat is the conversational hub of the research workflow and connects to the rest of the platform:
  • Document Search → find and gather the right filings, research, and news first, then analyze them in Chat (or collect them in a Data Room). Chat searches the same document library, filterable by type.
  • Data Room → keep a project’s documents together and point Chat at the room — or specific folders — for analysis grounded in your materials, with citations back to your documents.
  • Agent Studio → the same assistant, run autonomously on a task. Use Chat for live, interactive work; hand bigger, open-ended jobs to Agent Studio and come back to the results.
  • Grids → bring grid or table data into Chat as context to reason over; when you want the same questions answered consistently across many companies or documents, build a Grid as the structured complement to Chat’s narrative answers.
  • Reports → Chat can generate a sourced, multi-section report directly (optionally from a report template), and it also powers AI-assisted report editing — making changes section by section so you can see exactly what changed.
  • Skills → save your best instructions as reusable Skills and apply them in Chat for consistent results; the assistant also loads built-in analytical playbooks automatically.
What it takes in: your questions, attached files, @-mentioned companies/filings/data, Data Room folders, document-library selections, and connected accounts. What it hands off: cited answers, charts and tables, and downloadable PowerPoint/Excel/Word files and reports that can feed a deck, a report, or the next step in your workflow.

​
Example workflow: from a question to a sourced deck

A real multi-feature journey — “analyze this company and build a short investment summary deck”:
  1. Gather sources — use Document Search to find the latest filings, transcripts, and research on the company.
  2. Organize — collect the best documents into a Data Room folder for the project.
  3. Analyze in Chat — open Chat, set the sources to that Data Room folder plus standardized financial data, @-mention the ticker, and ask in Agent mode for the key drivers, valuation, and risks. Follow up to drill into anything thin.
  4. Build the model — in the same conversation, ask for a comparable-company or DCF model and download the Excel file.
  5. Produce the deck — ask for a summary presentation, pick a template style, and download the PowerPoint.
  6. Scale or formalize — for a consistent cross-company view, build a Grid; for a long-form write-up, have Chat generate a Report and refine it section by section.
The result: a sourced, calculated analysis and a finished deck, assembled in one conversation and across the platform.

​
Common questions

Speed vs. depth: Instant (single facts, fastest) → Balanced (everyday questions) → Agent (standard multi-step analysis, can build charts, models, and documents — the default) → Deep Research (most thorough, for big open-ended work, and takes the longest). Agent and Deep Research keep running if you close the tab; Instant and Balanced are meant to finish while you wait.
They use the same underlying assistant. Chat is the live, conversational way to use it — you ask, watch, and refine. Agent Studio runs it autonomously on a task so you can launch it and come back. For quick, interactive work, use Chat; for long hands-off research, use Agent Studio.
Use the sources controls to switch on or off your Data Room (or specific folders), the document library (by type), standardized financial data, the web, and a connected data warehouse if your organization has one. Narrow them for a focused answer; leave more on for broader coverage.
Yes — point it at a Data Room folder, attach files to your message (up to 12.5 MB each), or connect an account (cloud storage or email). You’ll see each attached item listed on your message.
Yes. Ask for the deliverable and download the file. For presentations you can choose a template style; for reports, you can start from a report template. It can also edit an existing PowerPoint, Word, or Excel file you provide.
Typing @ lets you mention a specific company and pin it to an exact fiscal year, quarter, and data type (fundamentals, a transcript, a filing, research, news, and more), so the assistant works from exactly what you mean.
Deeper modes do more thorough, multi-step work and may run several lines of research at once — that’s normal. Pick a lighter mode for speed; use Agent or Deep Research when depth matters.
Yes. Type a new instruction to steer a run in progress without starting over, or stop it entirely. Both are available while the assistant is working.
Yes — answers include inline citations. Click any one to open the supporting material and verify it.
Conversations are saved to your history, and Agent and Deep Research runs keep going in the background — reopen the conversation to pick up the result. Instant and Balanced answers are meant to complete while you’re on the page.
Chats are private to you, so there’s no built-in way to hand off a whole conversation. To share an answer, export it to PDF or Word, or email it.
Start a new conversation for a new topic. Your past chats stay in your history (searchable) until you delete them.

​
Getting help

For help using Chat or to talk through a research workflow, reach AllMind support through the in-app support option or your account team.