Zendesk AI Footnotes
What Zendesk's AI features do with your data, as Zendesk publishes it. Checked October 31, 2026. Zendesk in the Directory.
Agent access
Zendesk announced a Zendesk MCP Server on May 19, 2026, to connect Zendesk tickets, knowledge and other data to external AI systems, and said early access would start that summer. source, source 2, source 3 · checked October 31, 2026
Stage: announced.
Generative AI tools in the Suite (enhance writing, ticket summaries, admin copilot)
Zendesk Suite
| On by default? | Not published as of October 31, 2026 |
|---|---|
| How an admin turns it off | Each feature has its own setting; admins see the status of most AI features on the Overview page in Admin Center. source, source 2 · checked October 31, 2026 |
| Trains on customer data? | The third-party LLMs behind these generative features are not trained on Zendesk customer data, and LLM vendors that are Zendesk sub-processors cannot use Service Data to train their models. source, source 2 · checked October 31, 2026 |
| Retention | Zero data retention: LLMs run either on infrastructure platforms where the model provider never accesses Service Data, or on the provider's zero data retention endpoints. source, source 2 · checked October 31, 2026 |
| Where data is processed | LLM providers host no Service Data. Zendesk says it has access to OpenAI's European endpoints for European customers. source · checked October 31, 2026 |
| Model providers | LLMs hosted on AWS Bedrock, Microsoft Azure, Groq or Fireworks AI, and models used directly from OpenAI, Google Gemini and ElevenLabs. The sub-processor list (effective September 2, 2026 for new customers, October 2, 2026 for existing ones) names Google, Baseten, ElevenLabs, Fireworks.ai, Groq, OpenAI and Microsoft for generative AI. source, source 2, source 3, source 4 · checked October 31, 2026 |
| Price | Included in the Zendesk Suite at no additional cost. source · checked October 31, 2026 |
Copilot (auto assist, suggested first replies, call summaries, suggested assignees)
Zendesk Copilot add-on
| On by default? | Not published as of October 31, 2026 |
|---|---|
| How an admin turns it off | Copilot features are grouped in an AI hub, where admins can turn off any feature; each feature also has its own on/off setting. source · checked October 31, 2026 |
| Trains on customer data? | Not published as of October 31, 2026 |
| Retention | Zero data retention with LLM providers. source · checked October 31, 2026 |
| Where data is processed | Not published as of October 31, 2026 |
| Model providers | Auto assist uses an LLM; Zendesk does not name it. The sub-processor list names OpenAI and Microsoft, among others, for generative AI functionality. source, source 2, source 3 · checked October 31, 2026 |
| Price | Add-on listed at 50 USD per agent per month, paid yearly. The pricing page lists Suite Enterprise + Copilot as Talk to Sales. source, source 2, source 3 · checked October 31, 2026 |
AI agents
Zendesk Suite (Team and above) and Support (Team and above)
| On by default? | Not published as of October 31, 2026 |
|---|---|
| How an admin turns it off | Turn off an AI agent by unpublishing it: remove its channels (AI agent), its routing rules or email connection (AI agents - Advanced bought before March 10, 2026), or republish it without channels (legacy or essential AI agent). source · checked October 31, 2026 |
| Trains on customer data? | Not published as of October 31, 2026 |
| Retention | Not published as of October 31, 2026 |
| Where data is processed | AI agent data can be hosted in any Zendesk region (US, EEA, UK, Japan or Australia) for accounts created on or after March 10, 2026; older AI agents accounts support the EEA (Belgium) or US only. source, source 2 · checked October 31, 2026 |
| Model providers | Google Cloud is the cloud provider for AI agents, with data hosted in the United States or Belgium. For Forethought AI agents, the sub-processor list names Google, Mistral AI, Deepgram, Cartesia and Gladia. source, source 2, source 3, source 4, source 5, source 6 · checked October 31, 2026 |
| Price | Consumption, metered per automated resolution: a request the AI agent resolves without escalation to a human. The pricing page lists 5 included automated resolutions per agent per month on Support Team and Suite Team and 10 on Suite Professional (up to 10,000 a year), with more at 1.50 USD each committed or 2.00 USD pay-as-you-go. Zendesk's help article on this pricing model gives 15 a month on Enterprise. Zendesk is moving accounts, with 30 days' notice, to resolution allowances priced by resolution tier. source, source 2, source 3, source 4, source 5, source 6, source 7, source 8 · checked October 31, 2026 |
Connections to outside MCP servers
Generally available since August 10, 2026 on Suite Growth and above and Support Team and above. Admins connect public or private MCP servers in Admin Center (Apps and integrations > Connections > Connections) using OAuth over HTTP, then choose which tools action flows may use; auto assist, AI agents and custom agents (early access) can use those flows. Zendesk says external MCP connections may send data outside the account's selected region. source, source 2, source 3, source 4, source 5, source 6 · checked October 31, 2026
Compliance
- HIPAA: Suite Professional or Enterprise with Advanced Compliance and a signed BAA. The Copilot add-on is Healthcare Enabled; AI Agents - Advanced and Forethought AI agents are HIPAA Enabled add-ons. Early access programs are not covered. source, source 2, source 3, source 4 · checked October 31, 2026
- FedRAMP: Zendesk says it is FedRAMP authorized at Low Impact Software-as-a-Service (LI-SaaS) and lists AI Agents and the Copilot add-on in its FedRAMP Li-SaaS Tailored scope. Customers must use US-only data center locality and US phone numbers and must not store personal data in the services. The FedRAMP Marketplace lists the Zendesk Customer Support and Help Desk Platform as FedRAMP Certified, Class B (Low). source, source 2, source 3, source 4, source 5, source 6, source 7 · checked October 31, 2026
- ISO/IEC 42001: Zendesk lists Zendesk AI (automations and intelligence), Workforce Engagement Management and Zendesk AI Agents as in scope for ISO 42001. source · checked October 31, 2026
EU AI Act
Zendesk says it has reviewed the EU AI Act to prepare itself and customers to comply, and that its customer interaction and agent productivity features are generally not viewed as high-risk. source, source 2 · checked October 31, 2026
Not published by Zendesk as of October 31, 2026
- One default state for Zendesk AI features (Zendesk says some are on by default and others must be turned on, feature by feature)
- Whether Copilot and AI agents data trains Zendesk's own models, and whether that is opt-in or opt-out (Zendesk's pages differ: the AI Trust page says customers control it; the older AI Data Use page describes training shared models on sanitized data). The third-party LLMs are not trained on customer data.
- Retention period for AI agent conversation data
- Named LLM models used by Copilot and AI agents
- Which LLM providers serve Zendesk AI agents (the sub-processor list names LLM providers for generative AI functionality and Forethought AI agents)
- Prices of resolution tiers under the resolution allowance model
- Authentication method and current release stage of the Zendesk MCP Server after the May 2026 announcement
Changes
No changes recorded since the first check on October 31, 2026.
How this page is made, and how to send a correction: method.