- Your live setup — budgets, models, tools, storage, spend. These change by the hour, so the assistant queries them the moment you ask.
- Your firm’s documents — policies, handbooks, methodology. These are written and change rarely, so the assistant searches them and quotes the relevant passage with a citation.
Questions about your setup
Ask these in the conversation where you are already working, rather than leaving to hunt through a settings screen.In a project conversation
In a personal conversation
Everything from the project list works here too.
How to tell it looked something up
Above the answer you will see a line naming what the assistant is doing:Checking the project’s budget
Checking which models are available
Checking which tools are available here
Looking up your organisation’s spend
Looking up the project team
Why some questions are only answered in a personal conversation
This is the part that surprises people, so it is worth understanding rather than working around. A project conversation is shared. Everyone in the project sees it, and one answer goes to all of them. So in a project thread the assistant will only tell you things every member of that project can already see on screen for themselves. Your organization’s total spend across every client is not one of those things. So if you ask for it in a project thread, the assistant answers the part it can — this project’s budget — and declines the rest. It is not being obtuse, and asking again will not help: the figure would land in every member’s view and stay in the thread permanently. Ask it in your own personal conversation instead, where the audience is one person.Being on your own does not make you an admin
In a personal conversation the assistant still checks your role. If you are not an Org Admin, asking about organization-wide spend gets you your own usage for that period — never the firm’s. That is the same rule the admin screens already apply; the conversation is not a way around it.What the spend figures mean
Spend comes back split by who it is owed to, and deliberately never added up:- Your own provider keys — usage billed to you directly by Anthropic, OpenAI, Google or whoever holds the account.
- ParleHub-managed models — usage drawn from your prepaid credit.
If your organization has personal workspaces switched off, there is no personal conversation to ask in, so organization-wide spend cannot be reached this way at all. Use the Usage screen instead.
Questions about your firm’s documents
If an Org Admin has set up organization knowledge, you can ask about your firm’s written policies in the same breath:“What does our expenses policy say about booking flights for a client visit?"You will get the relevant passage, followed by a citation — the document name and version, as a link you can click to open the document itself. Three things worth knowing:
"What’s our standard notice period in an engagement letter?"
"Do we have a policy on using client data for training?”
- The citation is trustworthy by construction. Its label is resolved from ParleHub’s own record of the document, not written by the assistant — so it cannot cite one document and link to another, and it cannot invent one.
- Hover the citation to see the effective date of the version that was quoted.
- A retired policy cannot be quoted. When your firm supersedes a document, the old version comes out of the index — so you will not get a confidently-cited answer from a policy that no longer applies. In an older conversation, a citation to a since-retired document shows as
[document unavailable]rather than opening.