Company & terms

An assistant works better when it knows the company: what it does, what Pantarey is used for and which terms are specific to it. An administrator enters this once under Users → AI assistants → Company & terms. Every connected assistant receives it when connecting and can look it up at any time. The page applies to all users of the account.

The five entries

What does your company do? Industry, size, customers, products. One paragraph is enough.
What do you use Pantarey for? What is modelled here: invoices, leave requests, campaigns. Helps the assistant to place questions correctly.
Terms: Company-specific terms and their meaning, for example "Customer = organization in contact management". The assistant uses them to map what users say to data schemas and processes. Up to 50 entries, 80 characters for the term and 300 for the meaning each.
Rules for the assistant: Language, units, tone, what it must never do. For example "Answer in German. Amounts in euros. Summarize the input before starting a process."
Contact for Pantarey questions: Name and e-mail, optional. The assistant names this person when it cannot help.

The text fields hold up to 1,500 characters each, name and e-mail of the contact 200 each. Save applies the entries; the page shows who changed them last.

How the assistant uses it

It receives description and rules when connecting. Terms and details are fetched on demand via the tool get_company_info.
It knows that the information is maintained by hand and may be outdated or incomplete. If it contradicts the actual data, the data wins and the assistant points that out.
Rules are instructions to the assistant, not a security boundary. What it may really do is decided by roles, limits and switches.
Changes take effect on the next connection, after five minutes at the latest.

Example

The terms field says "Campaign = an order in the Orders data schema". Asked "Which campaigns are running for Meyer?", the assistant searches the Orders schema for Meyer, without asking what a campaign is.

Frequently asked questions

Where to start?

With the terms. They help most, because they avoid misunderstandings between everyday language and schema names. Then two or three short rules.

There are two similar data schemas, the assistant picks the wrong one.

A term clarifies this, for example "Contacts = schema Contact management (new); do not use the tpl_ template". Alternatively, the old schema gets separate rights for AI assistants under Permissions; the assistant then does not see it at all.

Do the entries apply to all users?

Yes, to every assistant connected to this account. Personal preferences belong in the rules of the respective assistant (such as the project instructions in Claude), not here.

Next step

AI in Pantarey: where AI works without a connected assistant, in service tasks and the designers.