AI Assistant

The AI Assistant is a per-business chat that has live context on the selected business. Use it to ask plain-language questions about ad performance, lead qualification, KPI trends, and pipeline activity without writing SQL or jumping between screens. Open it from the left sidebar under AI → AI Assistant.

​

Context

The assistant is always scoped to the business shown in the top-left Business Space selector — indicated by the Context: <Business name> chip in the top right of the chat (for example, Context: BNB Accelerator). Switch businesses from the selector and the chip updates; existing conversations stay with the business they were created against.

​

Chat layout

• + New Chat (top left) — Start a new conversation. Each chat is saved in the left rail under the business it belongs to.
• Conversation list — All prior chats for this business. Empty state shows No conversations yet.
• Chat Bot panel (right) — The active conversation. The empty state prompts How can I help you today? Ask me anything about your business data, campaigns, leads, and performance metrics.
• Message input (bottom) — Type and send. Conversations stream replies as they generate.

​

What you can ask

The AI Assistant is a marketing analytics assistant for your agency CRM and ad data. It answers in plain English by querying your database, SOPs, and lead records — not general chat. Scope to one business with the header selector, or ask portfolio-wide questions across all businesses.

​

1. Performance & metrics (counts and KPIs)

Direct numbers from your CRM and ad accounts.

​

2. Campaigns, ads & funnels

​

3. Lead profiles & CRM search

Questions about what leads said or their characteristics. These use semantic search over conversations and notes, so counts may be shown as “at least N”.

• “Which leads said they were full-time agents?”
• “Who answered yes to [qualification question]?”
• “Find leads with budget over $500K”
• “Show me details for contact [ID]”
• “Find leads similar to contact [ID]“

​

4. Coaching, SOPs & strategy (Knowledge Base)

Process and playbook questions from your uploaded agency docs.

• “What’s our pixel reset protocol?”
• “When should I kill an ad at MAX KPI vs Baseline KPI?”
• “What’s the launch checklist?”
• “How do we structure a sandbox campaign?”
• “Explain the Sandbox 2.0 testing framework”

​

5. SOP + live data combined (advanced)

The assistant can chain SOP knowledge with your account data.

• “Per our SOP’s MAX KPI rule, which campaigns last 7 days are over the line?”
• “What’s my cost per booked call last month vs the MAX KPI in the SOP?”
• “The SOP says no more than 5 active ads per adset — which adsets violate that?”
• “Audit my account against the SOP for the past 30 days — give me a prioritized action list”
• “Run a launch readiness check on my newest campaign using the SOP checklist”

​

6. Historical & diagnostic questions

​

7. Optimizer & recommendations

• Account analysis and KPI-based suggestions.
• Listing AI Optimizer recommendations.
• Metric comparisons across periods or businesses.

​

What it does not do well

• General knowledge unrelated to your marketing data or SOPs.
• Topics not in your knowledge base (for example, “What’s our refund policy?” if that policy isn’t documented).
• Guessing KPIs without looking up your account’s actual targets.

​

How questions are routed

The system classifies each question by complexity:

1. Simple — direct counts and lookups (“How many leads last week?”).
2. Analytical — comparisons, multi-step queries, lead-trait searches.
3. Deep — diagnostic and advisory questions (“Why did X happen?”, “What should we do?”).

Simpler questions return faster and cost fewer tokens; deep questions take longer and consume more.

​

Tips for better answers

• Use natural date phrases: last week, this month, past 30 days.
• Pick a business in the Business Space header selector for account-specific questions; leave it broad for portfolio comparisons.
• Be specific when you can — business name, funnel name, campaign name, and a date range produce the cleanest answers.

​

Tokens

The assistant uses your agency token balance (shown in the bottom-left of the sidebar). Heavier questions (long ranges, deep diagnosis) cost more tokens than simple lookups. Run a batch sync first if you suspect stale data — the assistant reads from the same tables.