Ask Intelligence
Ask a question. Get the figure, and its source.
Ask Intelligence is the ask box at the top of every page. Type a business question in plain words. An analyst reads your own data through read-only tools scoped to the brand you have open, and answers with the real number. When the data is not there, it says so, and points at the part of the product that would produce it.
rows read: search_keywords (41), content_plans (12) · brand: Nordvik Bearings
Illustrative. Brand, rows, and figures are samples.
Not a chatbot with your logo on it.
A general assistant answers a question about your business the way it would answer anyone: from what it remembers of the world, confidently, with no source. Ask Intelligence can’t, because it has nothing to remember. What it has is tools, and each tool returns rows from your account, filtered on the brand you have open.
So the answer to “what did we spend on this brand” is the figure in your ledger, broken down by action. The answer to “which clusters have no content yet” is a list of your clusters. And when the data can’t support the question, you get a sentence saying so. Not a plausible number.
Reads, never writes
Every tool is read-only. Nothing here publishes, launches, or spends. When an answer turns into work, it becomes a card on the board with an owner and a date, and that is a person’s decision.
One brand at a time
The scope is enforced in code before a query reaches the database. The model never chooses the scope; it cannot widen it.
Quotes the figure
Counts, volumes, budgets, spend. It never invents a keyword, a campaign, a figure, or a URL.
Says when it cannot
And says where the data would come from, which is usually a module you have not switched on yet.
The rule that matters most
It never invents a keyword, a campaign, a figure, or a URL. The instruction is written into the analyst, and the tools make it hard to break, because there’s no route from the question to anything outside the brand.
For a blank page with no brand data at all, use thePlayground.
Fifteen things it can read.
Each is a read-only tool, and each returns rows for one brand. The scope is enforced in code, as described on the security page. One exception sits outside the ring: the pipeline. Sales data belongs to the organization, so it is reachable only when the question is asked from Sales.
Four steps between the question and the figure.
Fifteen tools is a lot to hold in your head. You don’t have to. Here is what happens after you press enter.
You ask
As you’d ask a colleague
From the ask box in the header, on every page. In the language you think in. A follow-up refines the previous answer; the thread holds.
It decides
What to read
The analyst picks the tools the question needs and calls them, as many as it takes up to a fixed cap per run.
We run
The queries
Each query is filtered on the active brand before it reaches the database. The model never chooses the scope; it cannot widen it.
It answers
From what it read
A short answer first, then the supporting detail, in tables when tables help. Real figures from the rows it pulled. If the rows are not there, it says so.
The figure, and the rows it came from. Checkable by the person who asked.
The questions it gets before the meeting.
Most of them are the questions someone used to spend an hour assembling from four tabs before the meeting. Now the person who wants the answer asks for it, and spends the hour on what to do about it.
The same analyst also sits inside each module, so a question asked inside Search reads Search and nothing else.
rows read: search_keywords (41), content_plans (12) · brand: Nordvik Bearings
Illustrative. Brand, rows, and figures are samples.
Every answer says what it cost.
The billed cost of each answer comes back to your screen with the answer. The run is written to the organization’s ledger with your name, the brand, the model, and the exact amount, next to everything else the product did that day.
A run is capped at a fixed number of tool calls, so one question cannot run up a bill on its own. There is no seat fee. Usage is metered in tokens from the same wallet as the rest of the product.
Ledger line
WhoM. Lindqvist
BrandNordvik Bearings
Tools called3
Rows read171
Tokens6,420 in · 212 out
Billed to the wallet$0.0371
Sample line. The real one is under the answer.
Six things you can check.
Checked against the product on August 23, 2026. If one of these stops being true, the page changes.
It answers from whatever data you have added.
Switch on one module and Ask Intelligence starts reading it the same day. Add a second and the questions get more interesting. It comes with every account. The modules give it something to read.
No subscription. No seat fee. Money in a wallet, spent on tokens, storage and data, and your balance never expires.