# Zendesk AI agents — AI agents for customer service conversations and resolutions. > Source: AgentsIndex — https://agentsindex.ai/zendesk-ai-agents (structured, researched, re-verified) > Facts last verified: 2026-10-23 Zendesk AI agents automate customer service conversations across messaging, email, voice, and connected service environments. They use configured knowledge and policies to reason through requests, take actions in connected systems, and escalate when needed. | Fact | Value | | --- | --- | | Website | https://academy.zendesk.com/getting-started-with-ai-agents | | Price | from $55/seat month billed annually | | Free trial | Partly | 14 day free trial. No credit card required. ## What it costs to start **Free tier** · Partly — A time limited trial is available, but the product page does not describe a standing free AI agent plan. > 14-day free trial. No credit card required. > — https://www.zendesk.co.uk/service/ai/ai-agents/ · Hero trial callout **Self-serve signup** · Yes — The Suite Team plan containing AI agents has a public purchase path labeled Buy now. > Buy now > — https://www.zendesk.co.uk/pricing/ · Suite Team plan card **Card to start** · No — The published fourteen day trial states that no credit card is required before a customer starts it. > 14-day free trial. No credit card required. > — https://www.zendesk.co.uk/service/ai/ai-agents/ · Hero trial callout ## What it works with **Public API** · Yes — Zendesk publishes an AI Agents API for conversations, workflows, escalations, exports, and user data. > These APIs enable you to programmatically manage AI-powered conversations, ticket workflows, escalations to human agents, data exports, and user data. > — https://developer.zendesk.com/api-reference/ai-agents/introduction/ · Introduction, first paragraph ## What it does on its own **Acts on its own** · Yes — AI agents can reason through requests, take system actions, and complete resolutions without handoffs. > AI agents reason through multi-intent requests, ask clarifying questions and adapt in real time > — https://www.zendesk.co.uk/service/ai/ai-agents/ · Features, “Autonomously reason through complex workflows” **Limits on what it may do** · Yes — Administrators can govern agent operations by setting policies, reviewing behavior, and evaluating interactions. > Maintain full visibility and control over how AI agents operate – set policies, review behaviour and evaluate every interaction with built-in QA. > — https://www.zendesk.co.uk/service/ai/ai-agents/ · Benefits, “Govern with full control” **No code needed** · Yes — The course says administrators can create an agent in a few clicks without technical expertise. > deploy a fully functional AI agent in just a few clicks—no technical expertise needed! > — https://academy.zendesk.com/getting-started-with-ai-agents · About this course ## What happens to your data **Retention you control** · Partly — Customizable data-retention policies are documented only in the Advanced Data Privacy and Protection add-on. > The add-on includes capabilities for BYOK encryption, customisable data retention policies, data masking, PII redaction and access logs. > — https://www.zendesk.co.uk/trust-center/ · Product Security, Advanced Data Privacy and Protection **Data stays in your region** · Yes — Zendesk offers data-locality choices in the United States, Australia, Japan, and the European Economic Area. > Zendesk offers multiple data locality choices including the United States (US), Australia (AU), Japan (JP), or European Economic Area (EEA). > — https://www.zendesk.co.uk/trust-center/ · Cloud Security, Data Hosting Location **Take your data out** · Yes — The documented AI Agents API includes data exports, providing a programmatic route to retrieve data. > These APIs enable you to programmatically manage AI-powered conversations, ticket workflows, escalations to human agents, data exports, and user data. > — https://developer.zendesk.com/api-reference/ai-agents/introduction/ · Introduction, first paragraph ## Whether a company can buy it **Security audits** · Yes — Zendesk states that it is ISO 27001:2022 certified and publishes SOC 2 Type II reports under NDA. > Zendesk is ISO 27001:2022 certified. > — https://www.zendesk.co.uk/trust-center/ · Security Compliance table, ISO 27001:2022 **Single sign-on** · Yes — Zendesk supports enterprise single sign on using SAML or JWT for end users and agents. > subscribers can enable native Zendesk authentication, social media Single sign-on (SSO) (Facebook, Twitter, Google), and/or Enterprise SSO (SAML, JWT) for end user and/or agent authentication. > — https://www.zendesk.co.uk/trust-center/ · Product Security, Authentication Options **Audit log** · Partly — Audit Logs are available on Enterprise and Enterprise Plus accounts, rather than across every plan. > Zendesk offers Audit Logs to accounts with Enterprise/Enterprise Plus plans. > — https://www.zendesk.co.uk/trust-center/ · Additional Product Security Features, Audit Logs ## AI agents handle service work across channels Zendesk AI agents are part of its customer service platform rather than a standalone assistant. The product page says they work in messaging, email, voice, and other service environments. In messaging, the agent can clarify intent and complete complex resolutions within one conversation. In email, Zendesk says the agent interprets conversation history, applies policies, and executes multi step resolutions. The supplied Academy course frames the same capability as 24/7 automated support and covers knowledge replies, personas, escalation processes, and reporting metrics. The course is aimed at administrators, and its practical prerequisites include access to Zendesk Support admin settings, messaging, and Help Center. [Zendesk describes the channels](https://www.zendesk.co.uk/service/ai/ai-agents/) and [the course describes its audience and prerequisites](https://academy.zendesk.com/getting-started-with-ai-agents). ## Administration emphasizes knowledge, policies, and review The setup model starts with existing knowledge and policies. Zendesk says an administrator can connect the agent to its Help Center and external sources such as Google Drive or PDFs. Its stated control model includes setting policies, reviewing behavior, and evaluating every interaction with automated quality controls. That is useful for a service organization that needs agents to take actions across its systems, but it means the buyer remains responsible for the knowledge and policies that guide those actions. The Academy course says a learner should have Zendesk Support admin access and requires messaging and Help Center to set up a first agent. This is an administrative product workflow, not a general consumer chatbot. [The product page describes knowledge connections and controls](https://www.zendesk.co.uk/service/ai/ai-agents/) and [the Academy course names the setup prerequisites](https://academy.zendesk.com/getting-started-with-ai-agents). ## Public pricing places AI agents in Suite Team Zendesk places AI agents in the Suite Team plan. On the US pricing page, Suite Team is $55 per agent per month when paid yearly. The page also has a Buy now route for that plan, so a prospective customer can begin a self serve purchase. The AI agent product page separately advertises a fourteen day trial with no credit card required. That is a trial, not evidence of an ongoing free plan. Zendesk’s API documentation also says each account has an allowance of automated resolutions based on its plan, without giving the allowance on the cited introduction page. Buyers should obtain the current allowance and any extra resolution pricing before forecasting volume. [The pricing table identifies Suite Team](https://www.zendesk.com/pricing/) and [the API introduction explains resolution measurement](https://developer.zendesk.com/api-reference/ai-agents/introduction/). ## The agent API and webhooks support external connections There is a documented AI Agents API. Zendesk says it can programmatically manage AI based conversations, ticket workflows, human escalations, data exports, and user data. The API documentation cautions that these endpoints do not follow the general mechanisms and conventions of most Zendesk APIs, so implementation work should allow for that product specific behavior. Zendesk also publishes webhook documentation for sending an HTTP request to a configured URL when subscribed activity occurs in Support, Guide, Gather, or Messaging. The materials reviewed do not document an MCP server or an option to choose an underlying language model. [The AI Agents API introduction is here](https://developer.zendesk.com/api-reference/ai-agents/introduction/) and [the webhook guide describes event delivery](https://developer.zendesk.com/documentation/webhooks/). ## Enterprise controls add privacy, logs, and recovery terms Zendesk’s Trust Centre states that it is ISO 27001:2022 certified and says updated SOC 2 Type II reports are available on request under NDA. It offers data location choices in the United States, Australia, Japan, and the European Economic Area. Some controls are not included across every purchase path. The Advanced Data Privacy and Protection add on includes customizable retention policies, data masking, redaction, access logs, and BYOK encryption. Audit Logs are listed for Enterprise and Enterprise Plus accounts. Zendesk describes contractual recovery objectives in its Enhanced Disaster Recovery package, but the reviewed material does not publish an AI agent uptime commitment. [The Trust Centre documents the certifications, locality choices, add on, audit logs, and recovery package](https://www.zendesk.co.uk/trust-center/). ## What changed - 2026-10-24 — Price · Sections · What it does - 2026-10-23 — Free tier · Price · What it does ## Keeping it current These facts are re-checked against Zendesk AI agents's own pages and updated when they change. If you represent Zendesk AI agents, sign in and update the listing from your dashboard at https://agentsindex.ai/dashboard. The operations this site accepts from an agent are listed at https://agentsindex.ai/.well-known/agents.json — that catalog needs no key to read, and each entry states its own auth.