AI assistant
The AI assistant is available on every screen. It has two sides: it answers questions about your data in plain words, and — as an agent — it can run tools and make changes for you. Everything it does stays within your own permissions: it can't see or change anything you couldn't yourself.
Asking questions
Open the assistant and ask in plain words, for example "Show the top 10 ads by ROAS in the last 24 hours" or "Which geos lost money this week?". You don't write any query — the assistant fetches the data itself and answers with figures, tables and charts. Ready-made prompts at the start of a new chat give you a starting point.
It's context-aware: it knows the screen you're on. Open it on an ad and ask "why did this drop?"; open it while editing a rule or a script and it sees the code you're working on.
A chart the assistant draws can be pinned to a dashboard, so a question you asked once becomes a tile you return to. See My dashboards.
What the assistant can do for you
Beyond answering, the agent can act. These are the main things it can do:
- Read and analyse — pull any figures from your data, draw charts, and check a Britva rule against your ads before you save it.
- Ads — pause or resume an ad.
- Automation — create, edit, enable or delete a Britva rule; save or run a script.
- Tracker — create, edit or delete offers, networks, traffic sources and campaigns.
- Generation — generate an image (for a push banner, icon or elsewhere).
- Support — file a bug report for you.
Through its wider tool set (the same catalog exposed to the MCP server) it can reach more of the product — always within what your role allows.
Changes always ask first
The assistant separates reading from changing:
- Read-only questions — figures, charts, lists — run straight away.
- Any change — pausing an ad, creating a rule, editing an offer — is shown to you first: the assistant displays exactly what it's about to do, and it runs only after you approve. You can decline.
🎬 GIF: the agent proposing to pause an ad, and the approval card
So you can let the agent work freely without worrying it will change something behind your back.
Chats, models and cost
- Multiple chats — conversations are saved; you can rename or delete them and come back to an earlier thread.
- Model picker — pick which model answers, from the models your key allows.
- Cost — each answer shows what it cost; admins see the whole team's AI spend in AI spend.
- Insert code — when the assistant writes rule or script code, you can drop it straight into the editor you're in.
The assistant works step by step: for a bigger request it can run several tools in a row (fetch data, draw a chart, propose a change) within one answer — so you can ask for the outcome, not the individual steps.