PolyAI
by PolyAI
Enterprise voice AI agents for customer service, built on a proprietary dialog model
PolyAI is an enterprise conversational AI platform for building lifelike voice and chat agents that answer customer calls end to end. It is aimed at contact-centre, CX and operations leaders in hospitality, banking, utilities, retail and healthcare who need to automate high call volumes across many languages without degrading the caller experience.
PolyAI is a London-based company founded in 2017 and led by CEO and co-founder Nikola Mrkšić, which builds voice-first AI agents that handle customer service calls for large enterprises. Its Agentic Dialog Platform centres on Agent Studio, a no-code builder for CX and operations teams, and an Agent Development Kit (ADK) released on 22 April 2026 that lets developers build agents with coding assistants such as Claude Code and Cursor, with Git-based version control, code review, API keys and a CLI. Agents run by default on Raven, PolyAI's proprietary dialog model trained on more than a billion enterprise conversations, with GPT-5, Claude and Gemini available as alternatives; the company argues that putting agent behaviour in the model weights rather than in prompts makes it more stable under pressure. Wren is a natural-language builder inside Agent Studio that drafts a first version of an agent from a description. On 18 May 2026 PolyAI opened the platform, previously sold only to large enterprises, to any builder with two months free, and it now supports 75 languages across 25 countries. The company reports more than 100 enterprise customers and 2,000+ live deployments, naming Marriott, Caesars Entertainment, PG&E, UniCredit, Foot Locker, FedEx, Metro Bank and Fogo de Chão, and cites a Forrester Total Economic Impact study showing 391% ROI. In December 2025 it raised an $86M Series D co-led by Georgian, Hedosophia and Khosla Ventures, taking total funding past $200M. It competes with Sierra, Decagon, Parloa, Cognigy and the voice-agent offerings of CCaaS vendors.
A head of customer experience or contact-centre VP at a large, high-call-volume enterprise (hospitality, utilities, banking, retail) that wants to automate phone calls end to end.
Natural-sounding voice agents that resolve routine calls such as reservations, billing and account queries without a human, in dozens of languages.
At a Glance
- Category
- Audio & Voice
- Pricing
- Usage-based, Contact for pricing
- Target Market
- CIOs, Heads of Customer Experience, Contact Centre Leaders, Enterprise Developers
- Deployment
- Cloud-only, API-based
- Founded
- 2017
- Headquarters
- London, United Kingdom
- Customers
- 100+ enterprises, 2,000+ live deployments (Dec 2025)
Key Features
- ✓Raven dialog model
Proprietary model trained on over a billion enterprise conversations and used by default, designed to keep agent behaviour stable across long calls.
- ✓Agent Studio
No-code builder where CX and operations teams design, test, monitor and tune voice agents without engineering tickets.
- ✓Agent Development Kit (ADK)
Code-first toolkit launched April 2026 with API keys, CLI and Git workflows, letting developers build agents using Claude Code or Cursor.
- ✓Wren natural-language builder
Describe the agent in plain language and Wren drafts a first version, shortening setup to minutes according to the vendor.
- ✓Multi-model support
Swap Raven for GPT-5, Claude or Gemini per agent, avoiding lock-in to a single model provider.
- ✓75-language coverage
Agents operate in 75 languages across 25 countries, letting multinationals run one platform globally rather than regional point tools.
- ✓Analytics, guardrails and monitoring
Real-time analytics, safety filters and visibility into agent decisions help teams audit why an agent responded as it did.
Capabilities
Use Cases
- •Restaurant and hotel reservations
Voice agents take, change and cancel bookings around the clock; Fogo de Chão reports 95% guest satisfaction on automated calls.
- •Utility customer service
Agents handle billing, outage and account calls for utilities such as PG&E, freeing human agents for complex cases.
- •Banking contact-centre automation
Banks such as UniCredit and Metro Bank automate routine account queries by phone while routing sensitive cases to staff.
- •Healthcare scheduling
Direct Epic integration lets agents handle appointment scheduling calls for providers without staff re-keying data into the EHR.
Ideal For
Best For
- ✓High-volume inbound phone support in hospitality, restaurants and travel (reservations, changes, FAQs)
- ✓Utilities, banking and insurance contact centres automating account, billing and outage calls
- ✓Multinational enterprises that need the same voice agent in many languages and countries
- ✓Organisations that want both a no-code builder for CX teams and a code-first ADK for developers
Not Ideal For
- ✗Buyers who need to model cost before talking to sales: there is no published per-minute rate, minimum or rounding rule after the two-month free period
- ✗Deployments outside the US, UK and Canada that do not want to manage a separate Twilio account for telephony
- ✗Teams that primarily need text-chat support automation, where chat-first platforms may be a simpler fit
Integrations
Deployment
Market & Ratings
100+ enterprises, 2,000+ live deployments (Dec 2025)
Market Analysis
Pros
- ✓Large reference base of named enterprise deployments across hospitality, utilities and banking
- ✓Broad language coverage (75 languages, 25 countries) on a single platform
- ✓Broad compliance posture: SOC 2 Type II, ISO 27001, HIPAA, GDPR and PCI DSS
- ✓Model choice between Raven and GPT-5, Claude or Gemini
Cons
- ✗No published per-minute rate, minimum or rounding rule, so spend cannot be modelled without a sales process
- ✗Self-serve and enterprise run on different clusters, so realistic production testing still requires sales engagement
- ✗Wren builder usage is under an unpublished per-account fair-use allowance
- ✗ROI figures come from a vendor-commissioned Forrester study; Hacker News discussion is sparse and G2 reviews were not accessible
Pricing
Self-serve
Free for first two months, then unpublished
- ✓Agent Studio and Wren builder (fair-use allowance)
- ✓ADK with API keys and CLI
- ✓Telephony included in US, UK and Canada
Enterprise
Contact for pricing
- ✓Custom per-minute rate
- ✓99.9% uptime SLA
- ✓24/7/365 support and emergency line
- ✓Custom voices
- ✓Managed integrations and dashboards
Ongoing use is billed per minute of voice agent usage, but PolyAI publishes no rate card, minimum or rounding rule; self-serve is free for two months and then unpublished. Telephony outside the US, UK and Canada requires your own Twilio account, and custom voices, managed dashboards and integrations such as custom SIP or PCI Pal are enterprise-only.
Security & Compliance
Sources
This page was written from 9 sources, 5 on domains other than poly.ai.
- 1.poly.ai — poly.aivendor
- 2.poly.ai — pricingvendor
- 3.poly.ai — polyai raises 86m to transform how enterprises talk to theirvendor
- 4.poly.ai — polyai launches ADK to bring ai native development to enterpvendor
- 5.cmswire.com — polyai raises 86m series d for enterprise voice ai
- 6.cmswire.com — polyai makes enterprise conversational ai self serve
- 7.prnewswire.com — polyai opens its agentic dialog platform making the tech beh
- 8.cekura.ai — polyai pricing
- 9.hn.algolia.com — search
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Cartesia
Low-latency voice AI on state space models: Sonic text-to-speech, Ink speech-to-text and managed voice agents
Omilia Cloud Platform
Self-learning agentic voice AI for enterprise contact centers: voice agents, voice biometrics and fraud defense on one platform
Phonely
Voice agents running on Alma, a model trained on ten million real phone calls
Smallest.ai
Sub-second voice AI — small, specialised speech models for real-time enterprise voice agents