Conversational AI that answers from your live agency data — finance, ad-spend, tasks, projects, clients and more — with @entity mentions, role-focused assistants, and propose-then-confirm actions. Ask questions in natural language.
Ask questions about your agency in plain English. "How much did we spend on Meta for Client X last month?" or "Which tasks are overdue this week?" The AI understands your agency context — clients, projects, tasks, financial data — and retrieves the relevant information using a composite scoring system that combines semantic similarity, recency, importance, and entity matching.
Beyond pre-loaded context, the assistant can call dedicated tools to pull live numbers on demand — a finance snapshot, ad-spend pacing, open anomalies, project status, client overviews, social performance, briefs, and the knowledge base — so answers reflect what is true right now, with the exact sources it used one click away. Every tool runs inside your existing permissions, so the assistant can only ever see what you can. For anything that changes data, it follows a strict propose-then-confirm model: it drafts the action — for example, creating a task — and shows it to you for review; nothing is written until you confirm.
Ask which clients are most and least profitable — the assistant surfaces Agency Gross Income and delivery margin per client so you can see at a glance where the agency is making and losing money. You can also see how fast each client is burning its monthly retainer, with end-of-month projections so you know which accounts are at risk of over-running before they do. Flag where the agency is over-servicing — labour delivered beyond retainer scope — so account managers can have the right conversation early. And ask for a month-end revenue forecast or 90-day pipeline coverage against target, pulling from live Xero and project data.
Pick the assistant that fits the job — Finance, Marketing, Sales, or Account Management — or stay with the general Agency Assistant. Each persona keeps the same engine but leads with the right focus and narrows to the tools that matter for that role, always within your permissions. Switching is one click in the chat bar, and your choice sticks to the conversation, including on voice.
Type @ to mention specific clients, tasks, projects, or briefs in your AI conversation. Mentioned entities are pinned to the top of the AI context, ensuring the response is grounded in the exact data you are asking about. The autocomplete supports prefix-matching search across all entity types, so finding the right reference is fast even in agencies with hundreds of active projects.
Conversations persist across sessions with full history. Pin important conversations (up to 25) to your sidebar for quick access to frequently referenced topics. Start new conversations for fresh context or continue existing ones to build on previous analysis. The AI maintains awareness of the conversation history, so follow-up questions work naturally.
The AI chat uses Groq for inference, delivering responses with extremely low latency. Complex queries that involve multiple data lookups and analysis still return in seconds. The system uses edge-first intent classification (via Cloudflare Workers AI) to route queries efficiently, with Groq handling the conversational response generation for the best balance of speed and quality.