Back to all tools
PolyAI logo
Usage-basedAPI

PolyAI Review, Pricing & Alternatives (2026)

Enterprise agents for complex customer conversations across voice and chat

Last updated: October 26, 2026

What is PolyAI?

PolyAI is an enterprise conversational AI platform for companies handling customer calls, chat, and messaging. Its Dialog-RSN-1 API displays a price of $0.02 per connected minute, billed per second, while complete deployments are custom-priced. It is designed for regulated, high-volume support teams that need integrations, governance, and managed operations.

PolyAI website screenshot
PolyAI homepage, captured October 2026

PolyAI is an active enterprise conversational AI company founded in Cambridge in 2017 by Nikola Mrkšić, Shawn Wen, and Eddy Su. Its website reports 250 employees, more than 100 million annual calls, and over 2,000 live agent deployments. In December 2025, a Series D added $86 million, bringing total funding above $200 million, according to the company. Customers named by PolyAI include FedEx, Marriott, Caesars Entertainment, Volkswagen, and Nestlé. The platform targets customer-service conversations in which callers interrupt, change topics, and expect transactions or other tasks to be completed. Its published security controls include SOC 2 Type 2, ISO 27001, PCI DSS, GDPR support, regional hosting, and controls for HIPAA deployments. PolyAI offers voice, chat, and messaging agents for regulated, high-volume support operations requiring enterprise integrations and managed service.

A typical PolyAI workflow begins in Dialog Studio, where teams define an agent's purpose, policies, brand voice, tools, and escalation paths. Agent Builder supports no-code configuration, while the Agent Development Kit provides a CLI and Python package for engineering teams. Agents connect to CRM, reservation, payment, and other enterprise systems through prebuilt integrations, APIs, or MCP servers. Deployment can cover telephony, chat, and messaging, using a company's carrier or PolyAI numbers. Teams evaluate behavior with simulated conversations, test calls, and live-traffic A/B tests before monitoring latency and cost per agent in the console. Operational controls include human handoff, PII redaction, guardrails, and role-based access. The platform also supports regional hosting and integrations intended for regulated, high-volume customer-service operations, rather than only basic scripted call flows.

Raven is PolyAI’s proprietary customer-service model, trained on enterprise conversational data and described by AWS as post-trained on production conversations from more than 130 deployments. PolyAI reported Raven v2 in August 2025 and Raven v3.5 on March 10, 2026. Its training-data disclosure says Raven and Dialog-RSN-1 datasets contain hundreds of thousands of turns across tens of thousands of conversations, after redaction, filtering, and labeling, with some synthetic data included. Dialog-RSN-1, released in July 2026, is an audio-native model that combines turn-taking, speech recognition, function calling, and response generation for live customer calls. PolyAI reports roughly 280 milliseconds median first-token latency. Early access was available to new customers in July, while the public API is listed at $0.02 per connected minute, billed by connected time.

PolyAI bills ongoing agent use by connected time, calculated per second without rounding, and its Dialog-RSN-1 model-only API costs $0.02 per connected minute. The API permits two concurrent calls before a payment card is added and includes ten afterward; higher concurrency requires a team arrangement. Voice output is listed as free through November 19, 2026, while text output is metered by tokens. A complete deployment adds orchestration, telephony, enterprise integrations, monitoring, support, maintenance, and performance improvements, with pricing handled through sales. AWS Marketplace lists a 500,000-minute annual commitment for $175,000 and states that commitments are nonrefundable. PolyAI does not publish a general free tier or standard subscription price for the complete enterprise platform, making the API the clearest public entry point for smaller evaluations, while larger deployments require budget and procurement planning.

The tradeoff is procurement complexity and pricing designed for high call volumes. PolyAI is less suitable for solo creators than Retell AI, Vapi, or Twilio Voice, which offer faster self-serve access and more room for developer experimentation. PolyAI instead provides a managed customer-service platform for voice, chat, and messaging, with regional hosting, enterprise integrations, PII redaction, human escalation, and support for regulated deployments. Its Dialog-RSN-1 API is listed at $0.02 per connected minute, while complete deployments add telephony, orchestration, monitoring, maintenance, and support under custom pricing. PolyAI says customer data is not used for training without opt-in and is not sold, while its disclosure allows contractually authorized data for model development. Buyers should review retention, residency, and business-associate terms. Choose PolyAI when governed, reliable voice resolution matters more than a low-friction developer trial.

Who is PolyAI best for?

  • Large contact centers automating complex inbound calls.
  • Regulated companies needing regional hosting and audit controls.
  • Teams connecting voice agents to enterprise systems.
  • Businesses requiring managed deployment and human escalation.

PolyAI Pricing (2026)

Pricing model: Usage-based. All prices in USD.

PlanPriceWhat you get
Dialog-RSN-1 API$0.02/connected minuteModel-only access for custom voice stacks, billed by connected second. Two concurrent calls are available before adding a card, then ten are included.
Enterprise agent platformCustomOngoing agent use is priced by connected time and includes integrations, telephony, monitoring, maintenance, and support.
AWS Marketplace commitment$175,000/500,000 minutes per yearAWS lists a minimum annual commitment of 500,000 minutes. The listing says there are no refunds.

Pros and Cons

Pros

  • Per-second billing avoids rounding short calls to full minutes.
  • Dialog-RSN-1 is designed for audio-native live-call interactions.
  • One agent can run across voice, chat, and messaging.
  • Published controls include PII redaction and regional residency.

Cons

  • Complete platform pricing requires a sales conversation.
  • AWS Marketplace lists a $175,000 annual commitment.
  • API concurrency starts at two calls before adding a card.

Who should look elsewhere?

  • Hobbyists seeking a free voice-agent builder.
  • Teams requiring a self-serve monthly subscription.
  • Developers wanting a general-purpose conversational model.

Technical Details

Open SourceNo
Self-HostedNo
API AccessYes

Frequently Asked Questions

How much does PolyAI cost?

The Dialog-RSN-1 model-only API displays a price of $0.02 per connected minute, billed by the connected second. The complete enterprise platform is custom-priced and includes orchestration, integrations, telephony, monitoring, maintenance, and support. AWS Marketplace lists $175,000 for a 500,000-minute annual commitment. PolyAI does not publish one universal subscription price.

Does PolyAI have a free plan?

PolyAI does not list a free plan for its complete enterprise platform. Testing, simulated conversations, A/B tests, and monitoring are included platform capabilities, while ongoing agent use is usage-priced. The pricing page lists voice output as free until November 19, 2026, but connected minutes and text output can still incur charges.

How does PolyAI compare with Retell AI?

PolyAI centers on managed enterprise customer-service deployments, governance, integrations, regional hosting, and high call volumes. Retell AI offers a more developer-oriented route for prototyping voice agents. PolyAI publishes enterprise compliance controls, while its complete pricing requires sales engagement. Retell is generally the easier starting point for smaller teams and early experiments.

Does PolyAI train on customer data?

PolyAI says customer data is never used for training without opt-in and is never sold. Its training-data page says customer data may support Raven and Dialog-RSN-1 where contractually authorized and legally permitted, alongside licensed, owned, and synthetic data. PolyAI says personal information is redacted before model development. Customers should review retention and residency terms.

Can I use PolyAI commercially?

PolyAI offers commercial SaaS and API access under service agreements rather than downloadable model weights. Dialog-RSN-1 usage is billed by connected time. Raven and Dialog-RSN-1 are not presented as open-source releases. Commercial deployment therefore depends on PolyAI's contractual terms, including usage, data-processing, and support conditions.