May 19, 2026
The Chatbot Era Is Over: Zendesk Launches an Autonomous Service Workforce
At its annual Relate conference, Zendesk has replaced its deflection-focused bot model with a new approach built around specialised AI agents that are accountable for outcomes.
The company’s new vision, branded the Autonomous Service Workforce, puts a unified Resolution Platform at the centre of enterprise service operations. The platform draws on approximately 20 billion ticket interactions to continuously learn from every customer exchange, feeding insights back into automated responses through the Resolution Learning Loop.
Zendesk CEO Tom Eggemeier described the announcement as a generational turning point: “We are entering the age of the autonomous service workforce. We believe every business will soon run on specialised AI agents that work alongside human experts as one unified team.”
Building Custom Agents Without Code
One of the headline additions is Agent Builder, a no-code tool that lets organisations design, test, deploy, and improve AI agents tailored to their own workflows, policies, and business logic. The tool is aimed at teams that need to automate complex work across front, middle, and back office functions, without losing governance and oversight in the process.
Zendesk’s existing AI agents, built in part through its acquisition of Forethought, have also been upgraded. They now operate across messaging, email, voice, and large language model environments, carrying context and continuity across channels. Voice AI Agents have been expanded with multilingual support covering more than 60 languages, and can switch languages mid-conversation without losing context. This builds on Zendesk’s earlier voice and video service updates, which introduced natural-speech voice agents and screen sharing to help agents and customers resolve more complex issues together.
Employee Service Gets Its Own AI Agents
Zendesk has also extended its autonomous agent capabilities into employee support. Purpose-built AI agents for internal teams operate inside Slack and Microsoft Teams, pull answers from across enterprise systems, and enforce source-level permissions so employees only see information they are cleared to access.
This capability is powered by Zendesk’s acquisition of Unleash, the AI-powered enterprise search platform, to help employees find accurate answers quickly without creating security or governance risks, a problem Zendesk identified as a significant driver of internal support costs.
Copilots Across the Entire Team
Beyond fully autonomous agents, Zendesk announced expanded Copilot tools for human workers. Agent Copilot connects to internal and external sources from day one and take action on at least 30% of tickets without human input. Admin Copilot, already live, identifies operational issues and applies workflow changes in real time. Knowledge Copilot, now in early access, scans real customer conversations to flag outdated or missing content. Analyst Copilot, also in early access, helps teams identify trends and understand root causes through a new agentic analytics experience.
The platform also introduces Quality Score, which analyses 100% of both human and AI interactions on Suite Professional plans and above, giving service teams a continuous, objective view of performance quality.
Outcome-Based Pricing
Perhaps the most commercially significant element of the announcement is Zendesk’s expansion of outcome-based pricing. Under this model, every resolution charged to a customer is independently verified by a dedicated AI evaluation model. Spam and routine exchanges are excluded, and businesses pay only for interactions that are genuinely resolved end-to-end.
The model is a direct response to the industry’s longstanding problem with deflection metrics, where bots were rewarded for stopping customers from reaching a human agent, regardless of whether the actual problem was solved.
New Connectors and MCP Support
On the integration side, Zendesk announced a library of 40 prebuilt workflow connectors, covering services including Okta, Claude, and OneDrive, with more than 100 additional apps planned by year-end.
The company also announced support for Model Context Protocol (MCP), allowing AI agents and Agent Copilot to connect to external systems once and expand their capabilities automatically as new tools are added. A Zendesk MCP Server will let external AI systems connect to Zendesk tickets and knowledge in a governed way.
Futurum Research CEO Daniel Newman said the direction made sense as a system-level play: “To improve the experience meaningfully, AI has to be part of a broader system that can connect context, take action, and evolve with the needs of the business. That’s the kind of approach that can help organisations build a more scalable and responsive support experience over time.”
