Agents
Agents are how FunnelStory acts on what it knows about your accounts. Every agent runs on the Customer Intelligence Graph rather than on raw data, and follows the same shape: it detects a moment worth acting on, investigates the account, and acts — either drafting the next step for a person to approve (Assist) or acting directly (Auto).
Three kinds of agents
| What it does | You provide | Start here | |
|---|---|---|---|
| Outcome Agents | Works toward a revenue goal: learns from past renewals which adoption renews or expands, finds the accounts renewing soon that fall short, recommends next best actions, and measures daily who's moving | A product, a goal (Renewal or Expansion), and the windows to learn from and act on | Outcome Agents |
| Built-in agents | Does one job FunnelStory has built and tuned, such as investigating a needle mover and drafting the next step | A setting to turn it on | Built-in agents |
| Agents you build | Automates whatever your team needs: triggered by signals, needle movers, conversations, queries, or a schedule, with steps that query data, reason with an LLM, and act in Slack, email, or your CRM | A trigger and steps — on a canvas, or described to Claude in plain language | Getting started |
They work together. An Outcome Agent finds the at-risk accounts and the strategy; an agent you build works each of those accounts; the Outcome Agent measures whether they move.
Detect, investigate, act
| Stage | Draws on | Examples |
|---|---|---|
| Detect | Intelligence: signals, needle movers, predictions, conversations, cohorts | A champion leaves; a prediction moves to Churn; an account joins an at-risk cohort |
| Investigate | Wisdom: the Dossier, relationship history, facts with sources, what has worked before | Who else is engaged, what the account cares about, what's already open |
| Act | Your channels: Action Cards, Slack, email, CRM updates | Draft an email for a CSM to send; alert the account team; update a HubSpot property |
Assist or Auto. Anything a customer will read usually starts in Assist: the agent drafts an Action Card and a person reviews and sends it. Internal alerts can run in Auto from day one. See Design an Agent: Detect, Investigate, Act for a worksheet.
Related
- How FunnelStory Works — where agents sit in the architecture
- Built-in and Outcome Agents and AI Agents — the concepts
- Renari and the MCP Server — working with FunnelStory from an AI assistant
- Notifications overview — when built-in Slack or Teams notifications are enough