Wren dialog agent
Meet your new dialog partner
Wren helps you design, build, test, and continuously optimize every agentic dialog experience you deploy.
The agentic dialog loop
Proactively improve KPIs
Wren operates on a daily, adaptive loop to keep your agents performing at their best.
Self-improving
Wren analyzes real caller requests to find optimization opportunities that will actively improve your key metrics, like containment and CSAT.
Self-healing
Wren monitors live deployments to catch outages, bugs, or bad updates, and automatically drafts repairs.
The agent behind your agents
Wren makes everyone a forward-deployed engineer.
Actionable Daily Insights
Delivers daily performance reports and proactive recommendations for improving the customer experience.
Opportunities, ranked by impact
Analyzes real caller requests to find the changes that will move your key metrics, ranked by the metric you told Wren to prioritize.
From a description to a working agent
Describe the job and Wren builds the flows, prompts and tools behind it, ready to test before anything reaches a caller.
Wired into the systems you already run
Connects tools and data over MCP and REST, so your agent can read and write in the systems your team works in every day.
Test the change before a caller does
Replays a drafted fix against real conversations and simulated callers, so it is proven on the calls it was written for.
Repairs drafted the moment something breaks
Catches an outage, a failing integration or a bad update in live traffic and drafts the repair, ready for your team to review and release.
Every channel, every language, every agent you run
Wren supports your entire ecosystem, no matter how complex.
Multichannel
Provides a unified context across voice, web chat, messaging, email, and in-app experiences so customers never have to repeat themselves.
- Unified
- 5 channels
- Voice-first
Multilingual
Actively reviews and supports conversations in 45 live languages.
- Global
- 45 languages live
- 100+ locales
Multi-agent
Monitors every agent across your entire account simultaneously to ensure consistent quality.
- Whole estate
We co-build with you
Wren supports your entire engineering ecosystem, no matter how complex.
Built for mission-critical conversations
Complex service calls
Warranty intake end to end: identity verified, address confirmed, troubleshooting resolved, service request created. For your team: whole call types disappear from the queue.
Regulated & high-stakes
Banking identity verification, payments, healthcare intake, and outage lines at peak volume. For your customers: verified, secure, and resolved the first time.
Turn-taking-sensitive flows
Reference numbers, spellings, card digits. Turn-taking is a prompt, not a threshold. For your builders: no silence windows or probability thresholds left to tune.
Outbound & back office
Navigating phone menus and holds, human vs. machine detection, reactivating cold leads with CRM read and write. For your operation: work that used to need a person dialling.
Prioritize metrics that matter
Tell Wren to prioritize the metrics you care about most, like CSAT, containment, or booking rates. Wren automatically ranks its recommendations based on what will actively improve those numbers, rather than just reacting to the most recent errors.
Failures that cost you
Catch silent escalations, like a broken integration at 2 AM or a confusing prompt, before they drive up your cost-to-serve.
Revenue leaking quietly
Spot exactly where customers abandon flows or make unnecessary transfers. Wren finds these drop-offs in the call data and automatically drafts fixes to close the gap.
Drift you wouldn't otherwise see
Prevent metrics from slowly degrading as your call mix changes. By analyzing every single conversation, Wren catches performance drift while it is still a minor issue.
Meet the teams at the frontier of agentic dialog
“PolyAI has been a fantastic partner on our journey. The fact that we can handle a million calls a month, reliably, with a great experience for our members is a real key differentiator for us, and it allows us to continue to grow.”
— Ben Salter | Chief Product Officer and Co-Founder. Read more
“We wanted to provide a highly consistent, near instant voice service to customers that could answer a limitless number of FAQ’s.”
— Jo Lai, VP of Customer Experience, Hopper 
Get to know our platform
Wren works alongside the rest of the stack. Here is where to go next.
- PlatformExplore the studio
Dialog Studio
PolyAI's no-code builder for voice agents and conversational AI, and where Wren lives and works. Design and orchestrate every conversation: flows, tools, and brand voice, from no-code to full code without changing platform.
- Latest modelMeet the model
Dialog-RSN-1
The audio-native frontier model behind the conversations Wren builds. It reasons directly over raw call audio and responds in under 300ms in production.
- DevelopersRead the docs
ADK and APIs
Agents as code. Build and version with the Agent Development Kit, connect tools and data over MCP and REST, and run our models as an API endpoint. Everything Wren does, scriptable.
Latest updates
Former Rapid7 finance leader and technology investor brings 20 years of experience growing software businesses to scale frontier dialog for enterprises and builders.
The fastest reasoning model for real-time, complex dialog is available now on AWS SageMaker. Subscribe through AWS Marketplace and deploy it as a SageMaker endpoint inside your own AWS environment.
Voice runs on a clock that text never had. Here's why turn-taking breaks most voice agents, what Dialog-RSN-1 does differently, and which layers of the stack enterprises should own.
Frequently asked questions
What is Wren?
PolyAI Wren is the agentic loop for improving AI voice agents inside Dialog Studio. It builds agents from a description, reviews every production conversation on a schedule, raises what is broken or missing with the calls attached, drafts and tests the fix, and waits for a person to approve it. No code required at any step.
Is Wren part of Dialog Studio?
Yes. Wren works inside the studio you already build agents in, on the same platform and the same account. There is nothing separate to install, and nothing to move.
Does Wren change my live agent on its own?
No. Wren drafts the change and tests it, then waits. Every change to a live agent is reviewed and released by your team.
How do I monitor the quality of my AI voice agent?
Wren reviews every conversation and sends a daily report: what changed, what broke, and what to do about it, with the calls attached. Point it at the metrics you care about most, like CSAT, containment or booking rates, and it ranks what it finds by what will move them.
Does one customer's conversation data inform another customer's agent?
No. Your conversations, and everything Wren learns from them, stay inside your account. PolyAI never trains on your data without an opt-in, and never sells it.
How fast can we go live?
Weeks, not quarters. Wren builds a first agent from a description of the job, and our engineers co-build the first deployment with you — talk to a dialog engineer to scope your own timeline.