Reporting Agent queries your Power BI model and shows the query it ran. Assistant Avatar answers from your documents, with a citation — and with a face and a voice. Both are in production, both are pay as you go.
Both agents work in the same order: understand the question, choose the source, draw up a plan, run it and check the result before they answer. Only the source differs — for Reporting Agent it is your semantic model, for the Avatar your documents.
A question in Serbian or English becomes a query against your Power BI model. You get an explanation, a table that pastes straight into Excel and a chart — together with every step of the reasoning and the exact DAX that was run.
You choose a measure and a rule — “over the threshold” or “a change greater than X per cent”. The agent watches that measure with nobody attending it and notifies the people you choose when something moves out of line.
When something moves out of line, the agent puts a series of follow-up queries on its own and writes an explanation of the cause. Every figure in it is checked against the actual data.
You ask by voice or in writing; the avatar answers out loud, and at the same time the written answer stands on screen with citations to the specific documents.
No improvising and no “general knowledge” that your material does not support. The knowledge is organised into modules, and access to modules is granted per user.
The persona is configuration, not accident: the name, the appearance, the voice, how it introduces itself and what it does not discuss. If the voice is not set up correctly, the agent would rather not speak than speak in someone else’s voice.
The agent you trust is not the one that always has an answer, but the one whose answer you can check.
Reporting Agent shows the exact query it ran. The Avatar cites the document. You do not trust the answer because it is well written, but because you can check it.
“I do not know on the basis of your sources” is a feature, not a shortcoming. An agent that can say it does not know is the only one you can believe when it says it does.
No per-seat licences. You pay for what the agent actually did, with a monthly spend cap that you set yourself.
The whole team can ask — there are no per-seat licences. You pay for what the agent actually did, and you set the monthly spend cap yourself. No contractual commitment and no minimum term.
You are charged for a question answered and for a Pulse insight that was actually sent. You can run metric previews and test sends as often as you like, at no cost.
You are charged for an answer from the sources and for a minute of speech and avatar. Preparing your corpus — indexing the documents — is a one-off, priced by the volume of material.
You set the monthly spend cap yourself, so the cost cannot run away. The monthly report shows where it went, per user and per source.
A working session in which we point an agent at one of your Power BI models or one of your document corpora — so you see both the answers and the trail it leaves.
Request a quote