ChurnZero

More than seventy integrations with the business tools customers use daily feed a single intelligence layer that every agent reads, by the vendor's count. Classes named across first-party pages span CRM, support and ticketing, product usage and adoption telemetry, revenue data, and knowledge systems, with Zendesk and Confluence named specifically as connected knowledge sources. That is breadth across classes rather than depth inside one ecosystem. The count and the classes come from product pages rather than a connector list.

Sourcechurnzero.com/features/customer-success-ai and churnzero.com/customer-success-softwareread 2026-09-01

Multi-step execution is documented in two forms. The platform ships Journeys and Plays as automated sequences across onboarding, adoption, renewal and expansion, and the Pilot agent triggers AI actions inside that automation.

Separately, three Handoff agents carry an account across organizational stages, from sales to implementation to customer success and from one CSM to another, and Consult generates a full Success Plan that Compass then summarizes and sends. Work is chained across agents and stages rather than resolved in a single call.

The help center indexes journeys and segments, but no step, branching or failure handling documentation is published, including how a failed step is handled.

Sourcechurnzero.com/features/customer-success-ai and churnzero.com/customer-success-softwareread 2026-09-01

AI Knowledge Sources is a documented feature connecting agents to the customer's internal and external knowledge systems, Zendesk and Confluence named, which agents automatically reference as they analyze interactions and recommend next steps.

Alongside it the Customer Intelligence Profile holds product information, ICPs and personas, regulatory guardrails, segmentation, success metrics, pricing and packaging and competitive landscape, and live customer data arrives from the integration layer. The result is a maintained structure that persists and is read by every agent at runtime rather than assembled per request.

Sourcechurnzero.com/features/customer-success-ai and churnzero.com/features/ai-knowledge-sourcesread 2026-09-01

Three shipped controls rather than a governance claim. Per agent, the customer sets whether it suggests a next step or executes it, which accounts and contacts it is allowed to act on, and how often it runs. Several agents are drafting surfaces by construction, with Scribe producing contextual replies and Compass and Crux producing summary emails for a person to send. Custom Permission Groups bound who can configure any of it.

Sourcechurnzero.com/features/customer-success-airead 2026-09-01

Attestation and access controls are both documented first-party.

Attestation: the vendor's own trust center records a completed 2025 SOC 2 Type II examination with the customer letter available for download, an ISO 27001 annual review confirming the ISMS continues to meet the standard with an updated certificate published, and a third-party penetration test conducted in March 2026 with the engagement letter published; ISO 27001 certification is separately announced in the vendor's own 2024 press release and SOC 2 Type 2 in its own blog.

Control: the help center's admin documentation ships Custom Permission Groups, a ChurnZero User Hierarchy and Full Configuration Control, and the privacy policy states credentials are encrypted in transit. Third party listings that add PCI, FedRAMP or CSA STAR Level 1 are not backed by any of the vendor's own pages.

SourceChurnZero trust center, support.churnzero.com Managing Users and churnzero.com/global-privacy-policy; compliance.churnzero.comread 2026-09-01

No agent trace or audit surface is documented. The help center's six public categories, including a dedicated Admin Knowledge Base covering data, general settings, integrations and users, document no audit log, agent run history or activity trail. Reporting and health scoring are product output rather than a record of agent behavior.

Sourcesupport.churnzero.com category tree and churnzero.com/features/customer-success-airead 2026-09-01

State persists across sessions and agents read and write it: the Data Enrichment class maintains role assignments, gathers strategic account insights, identifies influencers and tracks signal changes over time, all written back to the account record where later runs and other agents read them, and the Customer Intelligence Profile is read by every agent at runtime.

No memory layer with a stated scope or lifetime is documented apart from the platform's own data model; the state lives in the account record, so deleting an agent's accumulated knowledge for one customer would mean deleting that customer's business data. How agent written account insights are reviewed and removed is not described.

Sourcechurnzero.com/features/customer-success-airead 2026-09-01

Multi-tenant cloud with no customer control over data location. The privacy policy states data may be stored and processed in the customer's region, in the United States, or in any other country where ChurnZero or its service providers operate facilities, and that the primary location is typically the customer's region or the United States, often with backup to a data center in another region. That describes the vendor's own operating practice, hedged with typically, not a region the customer selects.

The remainder of the apparatus is legal cover rather than residency: EU-US, UK Extension and Swiss-US Data Privacy Framework certification and standard contractual clauses govern lawful transfer, and the policy is explicit that transfers out of the EEA occur. No self hosted, on premises, VPC or private tenancy option is published, and whether data can be held in one region by contract is not stated.

Sourcechurnzero.com/global-privacy-policyread 2026-09-01

A self-serve AI Marketplace ships nineteen named agents in three classes, each with its own documented feature page: Workflow Assistance (Consult, Compass, Crux, Recap, Scribe, Verse, Retrospective, Pilot and three Handoff agents), Signal Detection (Beacon, Echo, Harbinger, Spotlight) and Data Enrichment (Archetype, Herald, Intel, Pulse, Vibes). The vendor states the customer controls which agents are used, which accounts and contacts they apply to and how often each runs.

Sourcechurnzero.com/features/customer-success-airead 2026-09-01

Three documented routes start an agent. Schedule: the customer sets how often each agent executes. Event: the Signal Detection class fires on account change, with Harbinger detecting relationship risk, Beacon spotting buying intent, Spotlight surfacing moments of impact and Vibes flagging signal shifts, and Journeys and Plays trigger on customer lifecycle events.

Request: agents also act on demand from within the workflow. On the outbound side the platform reaches customers through in-application digital engagement and walkthroughs and through email. These routes are named on product pages rather than in trigger configuration documentation.

Sourcechurnzero.com/features/customer-success-ai and churnzero.com/customer-success-softwareread 2026-09-01

The models behind the agents go unnamed, and customers have no selection.

The privacy policy has a dedicated Artificial Intelligence clause acknowledging generative AI features across the Products but naming no provider, and its only specific statement is that Google Workspace APIs are not used to train generalized models, which is a training restriction rather than a disclosure of who supplies the model.

No model selector, per agent model choice, bring your own key or routing policy is published. Access to ChurnZero AI inside Claude and ChatGPT points the other way, letting those assistants call ChurnZero rather than letting ChurnZero pick a model.

Sourcechurnzero.com/global-privacy-policy and churnzero.com/features/customer-success-airead 2026-09-01

Two inbound surfaces are documented. ChurnZero Connect is a Model Context Protocol server that carries live ChurnZero data into Claude and ChatGPT, mapping plain-language questions to fields and applying permission-based guardrails, so the customer's assistant calls the vendor. Separately the help center publishes an Implementation Resources category addressed to developers covering HTTP APIs, JavaScript snippets, batch uploads and integration setup. The integration catalog of more than seventy tools points outward instead.

Sourcechurnzero.com/features/customer-success-ai and support.churnzero.com Implementation Resourcesread 2026-09-01

Nothing the customer can test, debug or optimize is documented. The vendor states its AI is built and tested to solve customer success challenges, which is vendor-side quality work rather than something the customer runs. No sandbox, simulation, prompt or agent test harness, scored run, A/B comparison or output quality measurement is published, including in the help center's ChurnZero AI category. The Retrospective agent's confidence score on a churn classification is agent output, not a test the customer runs against the agent.

Sourcechurnzero.com/features/customer-success-ai and support.churnzero.com category treeread 2026-09-01

Agents act through documented programmatic routes rather than by operating software as a user. Every action described runs over the integration layer, the HTTP API or email, all of which are interfaces the vendor can call directly, and the in-application digital engagement and walkthrough tooling operates ChurnZero's own product surface rather than outside software. No browser session, virtual desktop or interface driving capability is claimed.

Sourcechurnzero.com/features/customer-success-ai and support.churnzero.com category treeread 2026-09-01