Zendesk announced Wednesday at the AI Summit a series of LLM-based products aimed at rebuilding dependence on human engineers.
At the heart of the new feature is an autonomous support agent that Zendesk believes can resolve 80% of support issues without human intervention. The system is complemented by a CoPilot agent that assists human technicians with the remaining 20% of problems, as well as management layer agents, voice-based agents, and analytics agents.
Shashi Upadhyay, president of products, engineering and AI at Zendesk, said the new agents are part of a broader shift in the support industry, as AI replaces many of the tasks traditionally performed by humans.
“The world will move from software built for human users to systems where AI actually does most of the work,” Upadhyay told TechCrunch.
Independent benchmarks suggest that modern AI models can take on the job. The TAU bench, designed to measure the model’s ability to call tools, includes a scenario in which the model must process returned products. This is very similar to many support tasks. The current leader, Claude Sonnet 4.5, solves 85% of the questions on the test.
After a chaotic investor battle in 2022, Zendesk made a series of AI acquisitions that laid the foundation for its current transition. The analytics agent being released today builds directly on the company’s Hyperarc acquisition, which was completed in July. Previous AI acquisitions include QA and agent service system Klaus (acquired in February 2024) and automation platform Ultimate (acquired in March of the following year).
Zendesk has been previewing the new system with existing customers, and Upadhyay said the results are promising.
tech crunch event
san francisco
|
October 27-29, 2025
“For our customers who are using this, we’re seeing a five to 10 point increase in consumer satisfaction,” he told TechCrunch.
Although rarely at the scale of Zendesk, large language models are often deployed for customer support. Companies from Airbnb to Regal Theaters are already experimenting with in-house chatbot support, many contracting directly with foundational model labs. However, these systems typically deal with information retrieval rather than more complex troubleshooting or taking autonomous actions.
If new initiatives in AI-based support are successful, the economic impact will be significant. Zendesk’s resolution platform already supports nearly 20,000 customers and resolves 4.6 billion tickets each year. In addition to Zendesk, the company employs 2.4 million customer service representatives in the United States and many more in other countries.
