How FunnelStory Works
FunnelStory turns your scattered customer data into a Customer Intelligence Graph — a context graph of computed intelligence and accrued wisdom about every account — and runs agents on it that act toward the outcomes you set. Your own systems stay the source of truth: FunnelStory reads from them, builds account context on top, and writes actions back out.
The path every answer and action takes is Data → Intelligence → Wisdom → Agents, delivered wherever your team works.
1. Data: connect and model your sources
FunnelStory brings data in two ways, and both keep it current on a schedule:
- Data models map data you define — a warehouse query, a CRM report — into FunnelStory's schema: accounts, users, activity, subscriptions, and more.
- Connections with built-in syncs, such as Gmail, Zendesk, or Gong, bring their data in directly, without a model to write.
Syncs pick up only what's new since the last run, and some sources push selected events as they happen. Everything is organized around the account, so data from every source lands on the same customer.
The data covers three kinds of sources:
- Structured data — product usage, CRM records, tickets, and billing and revenue data.
- Unstructured data — meetings, email, chat, documents, and notes.
- Third-party data — enrichment data, social signals, news, articles, and external intelligence feeds.
2. Intelligence: computed once, shared by everyone
On top of that data, FunnelStory continuously computes intelligence and stores the results. Because it's computed once, centrally, every person and every agent works from the same numbers — no one re-derives a health score in a spreadsheet or a prompt, and agents spend their effort reasoning instead of recomputing.
Out-of-the-box intelligence activates as your data is connected:
- Predictive scores — churn, neutral, or retention for every account, with the factors behind each.
- Needle movers — leading indicators of risk and opportunity found in what customers say and do, including Content Themes specific to your business.
- Relationship risk — single-threaded accounts and engagement drops.
- Customer journey — where each account sits in its lifecycle, through funnels.
- Historical engagement patterns — how each account has engaged with your product, support, and team over time.
- Temporal analysis — what behavior looks like in defined windows, such as the six months before a churn.
- Cohort analysis — how groups of accounts compare, and what separated the ones that renewed from the ones that didn't (see What's Next).
Custom intelligence extends the set: externally computed scores, datasets, and outputs from your own agents sit alongside the built-in intelligence and are available everywhere it is.
3. Wisdom: what FunnelStory learns about each account over time
Intelligence is point-in-time. Wisdom accrues: FunnelStory keeps a governed, growing understanding of every account that informs every decision, and it gets better as history builds up.
- Descriptive wisdom — what's true about the account. Who the stakeholders and champions are and what they care about, the account's concerns and use cases, how the relationship has changed, and what's been remembered from months or years back. It lives in the Account Dossier, relationship maps, and the account history. Every fact links to the conversation or note it came from, and your team can pin facts to keep them or dismiss ones that no longer apply.
- Decision wisdom — what to do about it. Which signals have actually preceded churn or expansion in your book, learned from your own outcomes: which needle-mover themes lift risk or opportunity, and which adoption levels renew better among comparable accounts. Outcome Agents turn this into cohorts and next best actions.
4. Agents: detect, investigate, act
Agents run on the graph rather than on raw data, which is what makes them reliable: every step starts from computed, cited facts. They follow one shape — detect a moment worth acting on (from intelligence), investigate the account (from wisdom), and act — either drafting the next step for a person to approve (Assist) or acting directly (Auto).
FunnelStory has three kinds:
- Outcome Agents — you set a goal, such as protecting renewals or growing expansion for a product; the agent finds the cohort of accounts that can reach it, recommends next best actions, and measures daily which accounts are moving.
- Built-in agents — agents FunnelStory ships for one job, such as investigating a needle mover and drafting an Action Card, on demand (Assist) or on a schedule (Auto).
- AI Agents you build — triggered by signals, needle movers, conversations, queries, or a schedule, and built on a canvas or described in plain language through vibe coding. They can post to Slack, send email, update your CRM, and call other systems.
Agents keep running in the background whether or not anyone is signed in.
Where you work
The same graph and agents reach your team wherever they already work:
- The FunnelStory app — Focus Areas, account pages, needle movers, Outcome Agents, and dashboards.
- AI assistants — Renari in the product, and Claude, ChatGPT, Cursor, or any MCP client through the MCP server, with skills for briefs, QBR decks, and reports.
- Slack, Microsoft Teams, and email — notifications, meeting briefings, and agent messages.
- Your CRM — CRM sync writes FunnelStory fields back to Salesforce and HubSpot, and the HubSpot app card shows account context on HubSpot records.
- Your own apps — Data Apps and the API.
Access and visibility
Every surface sees the same data under the same rules. Users see the accounts they're allowed to see — through assignments, shared teams, and CRM visibility, which mirrors account access from Salesforce and HubSpot — and the semantic query layer applies those rules to agents and AI assistants too, not just the app. See Roles and permissions and Security.
Related
- FunnelStory 101 — what FunnelStory does and where to start
- Customer Intelligence Graph — how the graph is built and queried
- Built-in and Outcome Agents and AI Agents
- Your First Week with FunnelStory
- Design an Agent: Detect, Investigate, Act