Zendesk Custom Agents powered by Claude reached one million executions in their first seven weeks of early access. Developed using Anthropic's technology, these agents automate complex service workflows unique to each business.
The milestone followed a four-month journey from proof of concept to early access. A five-person Zendesk team built the Custom Agent Builder using Claude Code and the Claude Agent SDK.
Early customer results are equally promising. According to Anthropic's Zendesk customer story, businesses running Custom Agents have reported up to 10 percent higher automated resolution and up to 80 percent lower human handle times.
These results reflect growing demand for specialized AI agents that complete the operational work behind customer service.
Key results of Zendesk Custom Agents built with Claude
One million executions in seven weeks: Zendesk Custom Agents powered by Claude reached this milestone during early access.
Four months from concept to early access: A five-person Zendesk team developed the Custom Agent Builder using Anthropic technology.
Higher automated resolution: Customers running Custom Agents reported improvements of up to 10 percent.
Reduced human handle time: Customers also reported reductions of up to 80 percent.
More complex work automated: Custom Agents execute business-specific workflows using company instructions, connected systems, and defined controls.
The challenge: Automating the work behind customer conversations
AI has already changed how businesses respond to customer questions. The next challenge involves the work required to resolve more complicated requests.
Consider a warranty claim that requires verifying a purchase, reviewing coverage, and coordinating a replacement. Or a billing dispute involving several systems and company-specific policies.
These workflows require employees to gather information, make decisions, and coordinate actions manually. The challenge grows when every business follows different processes.
Zendesk introduced Specialized Agents to address this complexity through two approaches. Industry AI agents provide prebuilt expertise for workflows common to particular industries, such as ecommerce returns and exchanges. Custom AI agents address the processes, policies, and exceptions unique to individual businesses.
Zendesk's work with Claude focuses on the second opportunity: Making business-specific automation more accessible without requiring companies to build every workflow from scratch.
Why Zendesk chose Claude for Custom Agents
Zendesk needed AI models capable of executing complex workflows accurately and efficiently. The team evaluated Anthropic and other providers using an internal testing system covering 15 metrics. The tests reflected the types of agents Zendesk customers would build.
Zendesk selected Claude Sonnet 4.6 for its performance on tasks such as interpreting documents and making decisions. The team also found that Claude required fewer planning steps for certain tasks, improving execution efficiency.
For businesses running agents at scale, the cost of completing a task matters alongside model pricing. Security and data locality influenced the architecture.
Zendesk runs Claude through Amazon Bedrock within its existing Amazon Web Services (AWS) environment. This keeps processing within Zendesk's established hosting regions and supports customer governance requirements.
The result is a foundation for Custom Agents that combine AI reasoning with business-specific instructions and operational controls.
From proof of concept to one million executions
Using Claude Code and the Claude Agent SDK, Zendesk's five-person team moved from proof of concept to early access in four months.
Customers build Custom Agents by writing instructions in plain English and connecting them to approved tools and business systems. These agents perform specialized tasks based on each organization's policies, processes, and operational needs.
One example highlighted in Anthropic's story involves incident management. When multiple customers report the same technical issue, a Custom Agent identifies related tickets, creates a Jira issue, and notifies engineers through Slack.
The agent updates affected customers while the team investigates. Once the problem is resolved, it closes related tickets and records an explanation. The workflow connects customer support, engineering, and communication without requiring employees to coordinate every step manually.
Within seven weeks of early access, customers had executed Zendesk Custom Agents one million times. The milestone demonstrated early adoption of agentic AI designed for the operational work behind service resolutions.
What customers are achieving with Custom Agents
The impact extends beyond the execution milestone.
According to Anthropic, customers running Zendesk Custom Agents have reported:
Up to 10% higher automated resolution, indicating more requests completed through AI.
Up to 80% lower human handle times, reducing the manual effort associated with service workflows.
Anthropic also highlighted a customer handling more than 14 million service requests annually. During a pilot, this organization achieved 99 percent accuracy in duplicate ticket merging using a Custom Agent.
This process typically requires employees to identify related requests, determine which tickets belong together, and consolidate them. Automating the work reduces repetitive administration and gives service teams more capacity for complex customer needs.
These results also reinforce the distinction between AI activity and customer outcomes. One million executions reflects usage, while automated resolution and handling time reveal operational impact.
This resolution-focused approach also underpins Zendesk's broader AI agents for customer service strategy. As businesses expand automation, they need to measure both the work completed and the quality of the resulting service.
Building confidence through control and visibility
Automating complex workflows requires businesses to trust the actions AI performs. Zendesk Custom Agents operate within defined permissions, allowing organizations to determine which actions an agent is authorized to execute.
Anthropic's case study also describes Zendesk's work on monitoring and testing capabilities, including tool-call logs, simulated conversations, and A/B testing. These capabilities give businesses greater visibility into agent behavior and opportunities to improve performance.
What's next for Zendesk Custom Agents powered by Claude?
The first million executions mark an early milestone in Zendesk's work with Anthropic.
The partnership demonstrates how specialized AI agents are moving deeper into customer service operations, executing tasks that previously required manual coordination or custom engineering.
This progress forms part of Zendesk's broader vision for Specialized Agents. Businesses can start with Industry Agents for common industry workflows and build Custom Agents around their unique requirements.
Together, these approaches extend automation across more of the work required to resolve customer problems. For businesses, the opportunity is to automate increasingly complex processes while preserving human expertise for situations requiring judgment and customer understanding.
Courtney Burry
Vice President of Product Marketing at Zendesk
Courtney is the Vice President of Product Marketing at Zendesk. She has worked in product and technical marketing roles for close to 30 years across a range of startups and Fortune 500 companies including National Semiconductor, VMware and Collibra. Prior to joining Zendesk, she was at Amplitude and spent three years leading product marketing, customer marketing and partner marketing. She loves building businesses and developing teams. She is also an avid vibe-coder and is incredibly passionate about AI.
Originally from Canada, Courtney is currently living in the Bay Area. She has also spent time living and working in Japan, Holland, the Philippines and New York.
