A new developer-first approach gives teams full control over building and continuously improving AI agents
PolyAI announced the release of the Agent Development Kit (ADK), a new way for teams to build, deploy, and improve agent AI for customer experiences.
ADK introduces an AI-native development model that embeds coding assistants at the core of building customer service agents. Instead of relying on static configuration and manual implementation, teams can generate, test, and evolve agents using the latest generation of coding assistants and development techniques.
Built for the way developers actually work
PolyAI’s ADK allows developer teams to:
- Work in your own environment with complete control, Use your favorite programming language and IDE with complete visibility into your codebase
- Accelerate development with AI coding assistants, Integrate tools like Cursor and Claude Code to generate and refine production-grade logic
- Build in minutes from any inputincluding diagrams, spreadsheets, and APIs.
- Manage agents like enterprise softwarewith version control, code reviews, and collaborative workflows.
“Most AI platforms for CX force developers to work within a UI, disconnecting them from how they actually build software,” said Sean Wen, co-founder and CTO of PolyAI. “With ADK, we’re changing that. Developers will be able to build AI agents using the same tools, workflows, and flexibility they use to build other critical systems.”
Also read: AiThority Interview with Glenn Jocher, Ultralytics Founder and CEO
From weeks of development to hours
The ADK brings a significant change in development speed and productivity to PolyAI’s enterprise customers. Enterprise customers using PolyAI’s ADK have achieved the following results:
- One of the world’s largest power companies built complex flows in hours that previously took weeks
- One of the UK’s largest banks processed hundreds of FAQs in minutes instead of hours of manual work.
By building directly against PolyAI’s platform and codebase, developers can also reuse proven production patterns for integration, logic, and behavior rather than starting from scratch. These patterns are informed by over 500,000 hours of real-world deployment work, built on PolyAI’s years of experience building and optimizing agents for the world’s most trusted enterprises.
“In addition to our customer success with ADK, we are also seeing a significant impact internally. More than 60% of our engineering work at PolyAI is now done autonomously through ADK-powered workflows, with output reviewed and guided by our developers,” Wen added. “Now, we are delivering the same capabilities to our customers, enabling more teams to build high-quality AI systems faster.”
A self-improvement system that learns from every interaction.
Beyond initial development, PolyAI’s ADK enables a self-learning loop.
By connecting to monitoring tools and automated QA pipelines, AI agents can identify performance gaps and continually self-improve.
This represents a major shift in the way agents are built for CX. That is, from static automation to systems that continuously learn and adapt without continuous manual effort.
A new standard for building enterprise-grade AI
While most CX AI vendors require developers to work within closed systems with little or no code access, version control, or integration with modern development tools, PolyAI’s ADK takes a fundamentally different approach that reflects how serious engineering organizations actually build.
Combining no-code and low-code capabilities already available in the company’s core Agent Studio platform, PolyAI is the only enterprise AI for CX solution that caters to both technical and non-technical users.
Also read: The infrastructure war behind the AI boom
