> ## Documentation Index
> Fetch the complete documentation index at: https://complete-docs-x17s80.allmind.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Grids — Complete Guide
> AI-powered research tables that answer the same questions consistently across many companies or documents — what they are, when and how to use them, the options, limits, and how they fit your workflow.
Grids are AI-powered research tables. Each **row** is something you want to analyze — a company or a
specific document (a filing, an earnings-call transcript, a research report) — and each **column** is
a question or a data point you want answered for every row. AllMind then fills each cell using the
same instruction across the whole table, so you get a consistent, side-by-side answer to the same
questions for dozens or hundreds of companies or documents at once. A grid turns work that would
otherwise be one-document-at-a-time reading into a single structured analysis you can sort, query,
and export.
*Find it under **Research Workspace → Grids**.*
## What it is
A grid works like a smart spreadsheet built for research:
* **Rows** are your subjects — companies (by ticker) or individual documents.
* **Columns** are what you want for each subject. A column is either a **data column** (a structured
market or financial figure pulled automatically — price, margins, multiples, estimates, ratings,
and so on) or an **AI column** (a plain-language instruction the AI answers for every row, such as
"summarize the key risks" or "what did management say about pricing?").
* **Cells** fill automatically. Data columns pull live numbers; AI columns read each row's document
(or fall back to live data and research) and write an answer grounded in what they find.
The value is **consistency at scale**: the same analytical framework applied identically to every
company or document, in minutes, instead of opening files one at a time.
## When to use it
Reach for a Grid when:
* You want to **compare many companies or documents on the same set of questions** — e.g., margins,
leverage, and key risks across the top gold miners, or revenue trends across a sector.
* You're doing **repetitive extraction** — pulling the same few figures or facts out of many filings
or transcripts.
* You want a **structured, sortable baseline** you can scan, filter, and export — not a single
narrative essay.
* You want the analysis to **stay current** as new filings arrive (see auto-refresh).
Use something else when:
* You need a **quick fact or a fast back-and-forth** → **Chat**.
* You want to **find or gather the source documents** first → **Document Search**, then collect them
in a **Data Room**.
* You want **one deep, open-ended narrative** on a single question ("analyze this company's
competitive position") → **Agent Studio**. A grid is the structured complement to that narrative.
* You want a **formatted, templated report or slide deck** as the final deliverable → **Reports** (a
grid's results can feed it).
* You just want to **look at one company's data** → the **Equities** data viewer.
## How to use it
There are two ways to build a grid: let the AI design it for you, or set it up yourself.
Open Grids and create a new one. Choose **Make Grid with AI** to describe what you want in plain
language, or build it manually by picking a grid type — a **Ticker grid** where each row is a
company, or a **Document grid** where each row is a document. You can place the new grid in a
folder right away.
For a **Ticker grid**, add companies — type them in individually, pull in a whole index or ETF,
use one of your saved **universes**, or name a sector or theme. For a **Document grid**, search a
company, choose a **document type** and a **time period** (or pick a ready-made preset like
"Earnings Transcripts (3Y)" or "Annual Reports (5Y)"), then add the matching documents. You can
also upload your own files — PDF, Word, Excel, PowerPoint, CSV, and more — from your computer,
Google Drive, OneDrive/SharePoint, Gmail, or Outlook.
Add **data columns** for structured figures (price, fundamentals, valuation multiples, estimates,
ratings, technicals, and more) by picking them from a catalog, and **AI columns** for anything you
want answered in plain language ("identify the key risk factors," "what changed versus last
quarter?") by writing a short prompt. You can also drop in ready-made column packs or apply a
saved template.
Start the fill. AllMind works through the cells **many at a time**, so the grid populates in a
fast, scattered burst rather than one row at a time — small grids finish in seconds, large ones
fill in waves over a few minutes. A progress strip shows how many cells are done, the percentage,
and an estimate of time remaining, and you can **Stop** at any point. You can also choose
**Generate Missing** (only the empty cells) or **Generate All**, or generate a single cell, row,
or column.
The grid opens in **Read** mode so you don't change anything by accident; switch to **Edit** to
make changes. **Regenerate** any cell, row, or column to redo it, edit a cell by hand, or use
**AI Edit** on a cell to revise it with an instruction ("make it shorter," "add detail on
revenue"). Use **Ask AI** to synthesize across the whole grid — "map the key risks quarter over
quarter," "which of these has the strongest margin trend?"
Save the grid's column structure as a **template** and its companies as a **universe** to reuse.
Sort and group rows, hide columns, or transpose the view to read it the way you want. **Export**
to Excel — download it or have it emailed — and set up **auto-refresh** if you want the grid to
keep itself current.
### Make Grid with AI
Describe the grid you want — by typing or by voice — for example *"compare the top 5 gold miners on
margins, leverage, and key risks."* The AI drafts a complete, **editable plan**: a grid name and
type, a proposed set of companies (or documents), and a set of columns, each tagged **data**, **AI**,
or **text**. Nothing is built until you approve it.
You review and adjust everything first: rename the grid, add or remove companies, dial an index up or
down, fix or drop columns, set how many documents to pull per company, and see a live estimate of the
total number of cells and how long the fill will take. Anything the AI couldn't confidently match
(an unrecognized company, a column with no library match) is flagged for you to fix or remove. When
you're happy, choose **Create & fill** and the grid builds and fills itself in the background.
"Make Grid with AI" selects companies for an index or theme by size (index weight), not by running a
quantitative screen. So "top 5 ... by margin" gives you real, relevant companies with the margin
column you asked for — but the *which-five* is by size, and you're expected to refine the row set in
review.
## Capabilities & key choices
### The two grid types
| Grid type | Each row is | Columns it uses | Best for |
| ----------------- | --------------------------------------- | ------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- |
| **Ticker grid** | A company | **Data columns only** (auto-filled figures) | A screener or comp sheet of live numbers — prices, fundamentals, multiples, estimates, ratings — across a universe or sector. |
| **Document grid** | A specific document (tied to a company) | **Both data and AI columns** | Reading and extracting the same answers from many particular filings or transcripts. |
You choose the type when you create the grid, and it's fixed for that grid. A Ticker grid is for
structured data, so its columns are all auto-filled data fields. A Document grid is for reading
documents, so it also lets you add AI columns that answer from each row's document. If you use **Make
Grid with AI** and your questions need to read documents, it builds a Document grid automatically.
### The two kinds of columns
* **Data columns ("auto-fill")** — structured market and financial figures, chosen from a catalog and
pulled automatically. They refresh live each time they run and carry their own context (the fiscal
year and period, the unit, whether a figure is an actual reported number or an estimate, and the
as-of date for prices). Numbers are formatted for you — dollars as millions/billions/trillions,
ratios as multiples, and so on.
* **AI columns** — a plain-language instruction you write, answered for every row. For a Document
grid, the AI answers from that row's own document; if there's no document (or no answer in it) and
the row has a company, it falls back to live data, current news, and the research library. AI
columns handle both **fact extraction** ("pull the reported revenue") and **open-ended analysis**
("five questions to ask management," "red flags," "what's missing"). There's no output-format
picker — you shape the answer through how you word the prompt.
* **Text columns** — blank columns you fill in yourself by hand (notes, your own labels). These are
never auto-filled.
### Ready-made column packs
Rather than writing every column from scratch, you can drop in curated packs of data columns,
including:
* **Stock data & fundamentals** — price and market data, income statement, balance sheet, cash flow.
* **Ratios** and **per-share data**.
* **Valuation multiples** — current and historical P/E, EV/EBITDA, EV/Revenue, P/B, and more.
* **Analyst estimates** — consensus EPS, revenue and EBITDA estimates, forward quarterly and annual,
revisions, and surprise (beat/miss) history.
* **Ratings & price targets** — analyst rating breakdowns, consensus, and target-price upside.
* **Metric changes** — period-over-period moves in key figures.
* **Technical indicators** — trend, momentum, volatility, and volume measures.
* **Sector performance & relative performance** — how a name is doing versus its sector.
* **Top / bottom performer checklists** — quick performance-ranking columns.
* **Risk statistics** — volatility, beta, Sharpe/Sortino, maximum drawdown.
* **ESG ratings**, **market sentiment** (a fear-and-greed gauge), and **macro indicators** (GDP, CPI,
unemployment, interest rates).
There are also ready-made **AI analysis templates** — pre-built question sets such as SWOT, Porter's
Five Forces, earnings analysis, risk analysis, and competitive analysis — that you add as AI columns
(Document grids only).
### Reuse: universes and templates
* A **universe** is a saved, named list of companies. Build one once and reuse it as the rows of any
future grid. You can create a universe by searching and picking companies, by uploading a list
(CSV, TSV, or Excel — common ticker formats from Bloomberg, Refinitiv, and Capital IQ are
recognized automatically), or directly from an existing grid's companies. Universe names are unique
to you, and universes are personal — they aren't shared. (There are no prebuilt universes; every
universe is one you create.)
* A **column template** is a saved, named set of columns (their prompts and data sources). Save the
structure of a grid you like and apply it to a fresh set of rows later. Column templates can be
**shared** with teammates and **copied**.
### Shape and explore the grid
Once a grid is filled you can make it readable: **search** across cell text and column names, **sort
and group** rows by company, document type, year, or quarter, **hide** columns or rows you don't need,
**transpose** the table to swap rows and columns, and auto-fit or compact the layout. The view you
arrange is the view that gets exported and the view **Ask AI** reasons over — what you see is what it
uses.
## What it can access & produce
**It can draw on:**
* **Your own documents** — filings, transcripts, research, investor-relations materials, and ESG
reports you search and attach to rows, plus files you upload from your computer, Google Drive,
OneDrive/SharePoint, Gmail, or Outlook.
* **Live market data** — prices (current and historical), daily changes, volume, ranges, and market
cap.
* **Fundamentals & ratios** — income statement, balance sheet, cash flow; segments and KPIs.
* **Valuation** — current and historical multiples and peer comparisons.
* **Estimates & analyst coverage** — consensus estimates, revisions, surprise history, ratings, and
price targets.
* **Corporate events** — earnings dates, investor days, conferences, and meetings.
* **Macro & market data** — official economic indicators, a fear-and-greed sentiment index, and live
index and yield quotes (these don't even need a company on the row).
* **News and the web** — recent company and sector news summaries and "what's driving the price"
analyses.
* **The research library** — broker and analyst research, regulatory filings (US and Canada),
earnings transcripts, press releases, IR materials, ESG reports, and sector data.
**It produces:**
* A **filled research table** — structured data and AI answers, consistent across every row.
* An **Excel export** of the grid you can download or send by email.
* A **saved universe** of the grid's companies and a **saved column template** of its structure, both
reusable.
* **Ask AI synthesis** across the whole grid — answers that read down and across the table.
Data cells carry source context (period, unit, estimate-vs-actual, as-of date), and AI answers are
grounded in the row's document — the AI is instructed to say the information isn't in the document
rather than guess.
## Tips & best practices
* **Be specific in your columns.** "Total reported revenue for the latest fiscal year" beats
"revenue." Precise instructions give cleaner, more consistent cells.
* **Match the grid type to the job.** Use a Document grid when you need answers pulled from specific
filings or transcripts; use a Ticker grid for cross-company data and metrics.
* **Start from a template or a pack.** Drop in a fundamentals or estimates pack, then add a couple of
custom AI columns for the judgment calls.
* **Keep a run to a sensible size.** Very large grids take longer; if you're analyzing hundreds of
companies with many AI columns, consider splitting into a few focused grids.
* **Refine surgically.** Tweak one column's instruction and regenerate just that column rather than
rebuilding the whole grid — only the affected cells recompute, and the rest of the grid is
untouched.
* **Save what works.** Turn a good column set into a template and a good company list into a universe
so the next analysis is one click.
* **Use Ask AI for the synthesis.** Let the grid give you the consistent baseline, then ask across it
for the narrative ("which names are deteriorating fastest, and why?").
## Limits & things to know
* **Per-run size guardrail (Make Grid with AI).** A single AI-built run is sized to keep results fast
and reliable: you'll see a heads-up around **1,000 cells**, and a run is capped at roughly
**10,000 total cells**, **2,500 AI-answered cells**, or **600 rows** — whichever comes first. If a
plan is bigger, you'll be asked to lower the company count, documents per company, or columns. You
can build larger grids manually, but the same practical ceilings apply to a single generation.
* **A plan you describe to the AI** can name up to 30 companies individually, pull from a few indexes
or saved universes (up to about 150 companies from one index, taken by size), and include up to 20
columns.
* **Uploads.** When you add your own files, you can upload up to **150 files at once**, up to **100 MB
each**.
* **How fast it fills.** Cells compute many at a time — dozens of AI answers and around a hundred data
lookups in flight at once — so grids fill in waves. A single generation run is time-bounded (on the
order of an hour or more for a very large grid); if it ever reaches that limit it stops and flags
the grid rather than hanging, and you can resume by generating the remaining cells.
* **If a cell can't be filled, it tells you** rather than going blank — you'll see a short message
like "Information not found in document" or a prompt to regenerate the cell.
* **"Top N" is by size, not a screen.** The AI builder ranks index and sector members by weight, not
by the metric you mention — refine the row set yourself.
* **AI columns read one row's document at a time.** A Document-grid AI cell answers from that row's
own document and won't pull facts from another row's document — by design, so companies don't bleed
together.
* **Sourcing today is provenance + grounding, not clickable footnotes.** Data cells know their period
and whether they're estimates or actuals; AI answers are grounded in the source document, and you
can open a row's document to check — but cells don't currently embed clickable citation links.
* **Your grids persist** until you delete them — grids, their cells, universes, templates, and folders
are saved to your workspace. Deleting a grid also removes its auto-refresh subscriptions.
## How it works with other features
Grids sit in the middle of the research workflow — they take in sources and companies and hand off
structured results:
* **Document Search → Grids.** Find the right filings, transcripts, and research, then attach them as
the rows (or the source documents) of a grid.
* **Data Room ↔ Grids.** A Data Room and a Document grid draw on the same kinds of materials
(filings, transcripts, uploads, connected sources). Gather a corpus in a Data Room, then build a
grid to extract the same fields from every document in it.
* **Agent Studio ↔ Grids.** An agent gives you depth on one question as a narrative; a grid gives you
the same answers consistently across many subjects. Use a grid to structure what several agent runs
explore — and use the grid's **Ask AI** to synthesize the table.
* **Grids → Reports.** Turn a grid's structured findings into a formatted, templated report or deck.
* **Grids ↔ Universes.** Saved universes feed a grid's rows; a grid can also spin its companies out
into a new universe for reuse elsewhere.
* **Chat.** For quick follow-up questions about anything in the grid.
**Keeping a grid current.** There are two ways to refresh:
* **Auto-update subscriptions.** Subscribe a grid to a company so it keeps itself current: when a new
matching document for that company appears — you choose which document types count (all, SEC
filings, Canadian/SEDAR filings, earnings transcripts, press releases, earnings slides, or
investor-relations categories) — the grid updates, either **adding a new row** for the new document
or **replacing an existing row** in place. Each subscription covers one company, so track several by
adding several. Updates are driven by new filings arriving, not a fixed daily or weekly schedule.
* **Check for new documents (manual).** Run an on-demand scan that checks every company already in the
grid for documents newer than what it holds, then pick which of the found documents to add.
**Sharing.** Grids and column templates can be shared with an individual or a whole team, with
**View**, **Edit**, or **Admin** access and an optional expiration date; recipients can be allowed to
download, export, or re-share. Anyone you share with can make their own independent **copy** into
their workspace. (Saved universes are personal — they aren't shared.)
## Example workflows
### 1. Compare a peer set on a consistent scorecard
**Goal: size up the major gold miners on the same metrics and risks.**
1. **Make Grid with AI** — describe it: "compare the major gold miners on margins, leverage,
valuation, and key risks."
2. **Review the plan** — confirm the companies (swap in any the builder missed), keep the margin,
leverage, and multiple **data columns**, and keep a "key risks" **AI column**. Create & fill.
3. **Scan and sort** — sort by leverage or valuation to see the outliers.
4. **Ask AI across the grid** — "which names screen cheap *and* have improving margins?"
5. **Save** the column set as a **template** and the companies as a **universe** for next quarter.
### 2. Read every transcript the same way (sector deep-dive into a deck)
**Goal: analyze the gold sector and build a presentation.**
1. **Document Search** — find the latest earnings-call transcripts and filings across the sector.
2. **Data Room** — collect the best sources into a room for the project.
3. **Document grid** — build a grid with one row per transcript and AI columns for the same questions
("management's tone on costs," "guidance changes," "key risks," "five questions to ask
management"). Generate, so every transcript is read the same way.
4. **Agent Studio** — run a deeper narrative on the one or two themes the grid surfaces as most
important.
5. **Reports** — combine the grid's structured findings and the agent's narrative into a presentation
from a template.
The result: a consistent, sourced analysis across the whole sector plus a finished deck — assembled
from across the platform.
### 3. A self-updating monitoring grid
**Goal: keep an eye on a watchlist as results come in.**
1. Build a **Ticker grid** from a saved **universe** (your watchlist) with data columns for valuation,
estimates, and ratings.
2. Add an **auto-update subscription** per company for earnings transcripts, set to **add a new row**
when a new transcript lands.
3. Each earnings season the grid grows itself; open it, run **Check for new documents** if you want to
pull anything in on demand, and **Ask AI** to summarize what changed.
## Common questions
In a Ticker grid each row is a company and the columns are auto-filled data fields (prices,
fundamentals, estimates, ratings). In a Document grid each row is a specific document (a filing or
transcript) tied to a company, and you can also add AI columns that answer *from that document*.
Use a Document grid when you want to read and extract from particular documents; use a Ticker grid
for cross-company data and metrics. The grid type is set when you create it.
A data column pulls a structured figure (price, margin, multiple, estimate, rating) automatically
and refreshes it live. An AI column answers a plain-language instruction for every row — reading
the row's document, or falling back to live data and research when there's no document. Text
columns are blank columns you fill in yourself. (Ticker grids use data columns only.)
Use **Make Grid with AI**: describe what you want, review the proposed companies and columns,
adjust, and create. You always get to edit the plan before anything runs.
Large grids fill in waves — many cells compute at once, but hundreds or thousands of cells still
take time. Small grids finish in seconds. A single run is time-bounded (about an hour or more for
a very large grid); if you need more, split it into a few focused grids.
When you build with AI there's a per-run guardrail — roughly up to 10,000 cells, 2,500
AI-answered cells, or 600 rows, with a heads-up around 1,000 cells. You can build larger grids
manually, but very large ones take longer to fill.
Yes — regenerate a single cell, a row, or a column, and only those cells recompute. The rest of
the grid stays as-is. You can also use **AI Edit** on one cell to revise it with an instruction
("make it shorter"), or edit any cell by hand.
Cells already finished stay filled; not-yet-started cells are skipped. You can pick up where it
left off by generating the remaining (missing) cells.
Yes — subscribe it to a company and choose which document types should trigger an update. When a
new matching filing arrives, the grid adds a new row (or replaces one). It's driven by new filings,
not a clock. You can also run **Check for new documents** any time to pull in newer documents on
demand.
Yes — export the grid to a formatted Excel spreadsheet: download it directly, or have it sent as
an attachment by email.
Save its column structure as a template and its companies as a universe. Templates can also be
shared with teammates; universes are personal to you.
Data cells carry their context (period, unit, estimate vs. actual, as-of date), and AI answers are
grounded in the row's document — which you can open to check. Cells don't currently include
clickable citation links.
Chat is for quick cited Q\&A; Agent Studio is for one deep, open-ended narrative. A grid is for
answering the *same* questions consistently across *many* companies or documents — the structured
complement to both.
## Getting help
For help building a grid or designing a research workflow around one, reach AllMind support through
the in-app support option or your account team.