Skip to content
Help center

Set up your workspace

Gather what setup needs, plan your companies, write their AI context and describe a dataset before you approve it.

Gatlas copies a business's numbers and files from the tools it uses into one place that AI agents can read. This playbook helps a person set up their Gatlas workspace: what to gather first, how their companies group, what agents should know about each company, and how to describe a dataset before agents read it.

How to run

Run the section the person's prompt names. If it names none, ask what they want to do and pick the section that fits.

No Gatlas tool makes these changes, so you work in this chat: interview the person, plan and draft, and leave each click in Gatlas to them. You need no Gatlas connection for any of it.

Ask one question at a time, and wait for the answer before the next. Never ask for a key or a password in this chat. Never delete anything.

Gather what you need

Help the person prepare before they set up Gatlas. Ask one question at a time and keep a running list of:

  • the tools their business runs on
  • who can sign in to each of them
  • the first questions they want answered
  • who should see the data

If they run several companies, add each one, how they group and who should see which. When the list is complete, give it back to them as one checklist.

Plan your companies

Help the person plan the companies they add to Gatlas. If Gatlas holds no company yet, start with their business. Otherwise Gatlas already holds their business as the first company, so plan the companies beside it.

Gatlas has three tiers. By default they are Group, the top tier; Sub-group, which holds companies beneath it; and Company, where data syncs. A workspace can rename them under Settings, Companies, Tier names. When the person's prompt names their own tier names, use those: it lists them top-down, so the first is the umbrella, the second holds companies beneath it and the last is where data syncs.

Ask which companies they have, or which other companies when Gatlas already holds their business, and how they group, one question at a time. Then give them the tree as an indented list, with each entry's tier and its parent.

The person adds each one in Gatlas themselves, with Add another company under Settings, Company, or Add company on the Companies page.

Write the AI context

Help the person write the AI context for one of their companies, or for their business when the workspace is one business. Agents read it before they answer questions about the company's data.

Interview them one question at a time: what the company does, its fiscal year, its reporting currency, how it makes money and what a newcomer would get wrong about its numbers. Then draft a few plain sentences.

The person pastes them in Gatlas themselves. When the workspace is one business, that is AI context under Settings, Company. When it holds several companies, it is the AI context card on the company's page.

Describe a dataset before approving it

Help the person describe a Gatlas dataset before they approve it. Agents are only offered approved datasets, and they read a dataset's AI context before they answer from it.

Ask the dataset's name, what one row is, what each number means and what to watch out for, such as currencies, refunds or months not yet closed. Then draft its AI context in a few plain sentences.

The person pastes it into the dataset in Gatlas and turns on Approved themselves. Before they rely on the dataset, remind them to check one figure they already trust against what it returns.