
PolyAI
The pre-LLM-era enterprise voice AI incumbent — $750M-valued, 200+ enterprise customers — that opened a genuine self-serve, free-to-start tier in May 2026.
Best forTeams that want a proven, enterprise-scale, compliance-certified voice AI vendor and are willing to start on its new self-serve free tier before deciding whether a larger, sales-led enterprise engagement makes sense.
Our verdict
PolyAI is the pre-LLM-era incumbent of the voice-agent category — founded in 2017, years before Vapi, Retell AI, Bland AI or ElevenLabs Conversational AI existed — with real enterprise scale to show for it: 200+ customers, thousands of live deployments, a $750M valuation, and a Forrester study (vendor-commissioned, but methodologically disclosed) modeling 391% three-year ROI for a composite enterprise customer.
The pre-LLM-era enterprise incumbent that opened a real self-serve, free-to-start tier in May 2026 — but what happens after the free period is still undisclosed.
The Agents Index digest
New agents, fresh verdicts, and who's earning the top spots. One short email.

What's great
- Deep enterprise proof: 200+ named customers, thousands of live deployments across 75+ languages in 25+ countries, and a $750M valuation (December 2025 Series D) with NVIDIA’s NVentures among the backers.
- A genuine self-serve tier as of May 2026 — free for two months from just an email address, via a no-code Poly Agent Builder or a real developer API (ADK with self-serve keys) — a real change from its previous sales-only-only access.
- A vendor-commissioned but methodologically disclosed Forrester Total Economic Impact study modeling 391% three-year ROI ($11.3M NPV, under-6-month payback) for a composite enterprise customer, including named, quantified benefit categories rather than one bare percentage.
Watch-outs
- What happens after the 2-month self-serve free period is not publicly disclosed anywhere — no self-serve rate card, so budgeting past month two means either guessing or starting a sales conversation.
- PolyAI’s own reported GAAP financials show operating losses widening to $26.6M (from $23.8M) against $15M revenue for the 12 months to January 31, 2025 — spending well ahead of reported revenue even as forward ARR guidance (a less-audited metric) is projected near $40M.
- The Forrester ROI figures come from a PolyAI-commissioned study interviewing just 4 customers combined into one hypothetical composite organization — directionally useful, but not a broad independent sample, and its 5.0/5 G2 rating rests on a thin sample of only about 12 reviews.
Also note: The self-serve free tier is time-boxed to two months with no published rate for after that — real budgeting still requires either waiting to be billed or a sales conversation. · Enterprise-scale deployments (the kind PolyAI is best proven at) still run through a separate, custom-quoted, sales-led engagement rather than the new self-serve on-ramp.
What is PolyAI?
PolyAI is a London-founded (2017) enterprise voice AI company that grew out of Cambridge machine-learning research, built by co-founders Nikola Mrkšić (CEO), Shawn Wen and Eddy Su. It raised an $86M Series D in December 2025 (co-led by Georgian, Hedosophia and Khosla Ventures, with NVIDIA’s NVentures also participating) at a $750M valuation — up from $500M in May 2024 — pushing total funding past $200M. Its Agentic Dialog Platform, built on the proprietary Raven dialog model, now also supports GPT-5, Claude and Gemini as reasoning engines and runs thousands of live phone and chat voice-agent deployments across 75+ languages in 25+ countries for 200+ enterprise customers, including PG&E, UniCredit, Caesars Entertainment, Marriott and FedEx. Until May 2026, access was enterprise, sales-led only; PolyAI then opened a genuine self-serve tier — free for the first two months, sign-up requiring only an email address.
What does PolyAI do?
PolyAI’s Agentic Dialog Platform is built on Raven, its proprietary dialog model trained on over a billion enterprise conversations, and — since a May 2026 platform opening — also supports third-party reasoning models (GPT-5, Claude, Gemini) alongside Raven. Two build paths now exist: Poly Agent Builder, a no-code, natural-language interface that configures an agent, its knowledge base and conversation flows automatically (PolyAI claims production-ready results in under 10 minutes), and the Agent Development Kit (ADK) — a CLI and Python package with self-serve API keys, native integrations and a Git-based workflow for engineers building from their own IDE. A shareable testing environment lets a team validate agent behavior across channels before going live. Deployed agents handle account management, authentication, call routing, billing, booking, order management, troubleshooting and FAQ resolution across phone and chat, integrating with 130+ telephony, CRM, knowledge-base and productivity systems via a self-serve Connect Portal (managed OAuth 2.0 with automatic health-checks). SOC 2, HIPAA, GDPR and PCI DSS compliance ship with the platform. The self-serve tier is free for the first two months from just an email address; PolyAI has not disclosed what pricing applies after that period, and its largest enterprise deployments (like PG&E’s, handling 16M calls/year) still run on a separate, custom-quoted, sales-led engagement rather than the self-serve on-ramp.
Key features
- Raven dialog model, now multi-model: PolyAI’s proprietary Raven model, trained on 1B+ enterprise conversations, now runs alongside third-party reasoning models (GPT-5, Claude, Gemini) since the May 2026 platform opening.
- Poly Agent Builder (no-code): Describe a use case in natural language and the builder configures the agent, knowledge base and conversation flows automatically — production-ready in a claimed under 10 minutes.
- Agent Development Kit (ADK), self-serve: A CLI and Python package with self-serve API keys, native integrations and a Git-based workflow — build, version and deploy agents from your own IDE without a sales call.
- Connect Portal integrations: 130+ pre-built telephony, CRM, knowledge-base and productivity integrations via a self-serve portal with managed OAuth 2.0 and automatic health-checks.
What are PolyAI's use cases?
- High-volume regulated-industry phone support: A utility or bank routes account inquiries, fraud alerts or outage calls through a PolyAI agent to absorb extreme call-volume spikes — PG&E’s deployment handles 16M calls/year and tens of thousands per day during emergencies, with 67% containment and 35,000 labor hours saved.
- Self-serve prototype before an enterprise commitment: A team signs up with just an email, uses Poly Agent Builder to configure a working agent in minutes on the free two-month tier, and evaluates it before deciding whether to pursue a larger, sales-led enterprise deployment.
- Developer-built custom voice agent: An engineering team uses the self-serve ADK to build and deploy a voice agent from their own IDE, wiring in one of 130+ Connect Portal integrations rather than working through PolyAI’s enterprise sales process.
What does PolyAI integrate with?
- Telephony/voice routing, CRM & knowledge, productivity/collaboration and vertical systems — 130+ pre-built connectors via the self-serve Connect Portal (managed OAuth 2.0, automatic health-checks)
- Reasoning models — the proprietary Raven model, or GPT-5, Claude and Gemini as of the May 2026 platform opening
- Developer tooling — self-serve Agent Development Kit (CLI + Python), Git-based local development workflow
Why use PolyAI?
- A genuine enterprise track record: 200+ named customers, thousands of live deployments across 75+ languages and 25+ countries, and a $750M valuation backed in part by NVIDIA’s NVentures.
- A real self-serve on-ramp as of May 2026 — free for two months from just an email address, with both a no-code builder and a genuine developer API (ADK), where previously every engagement required a sales call.
- A vendor-commissioned but methodologically transparent Forrester study modeling 391% three-year ROI, $10.3M in labor-cost savings and under-6-month payback for a composite enterprise customer — specific, checkable figures rather than a bare marketing claim.
Pros & cons
Pros
- Deep enterprise proof: 200+ named customers, thousands of live deployments across 75+ languages in 25+ countries, and a $750M valuation (December 2025 Series D) with NVIDIA’s NVentures among the backers.
- A genuine self-serve tier as of May 2026 — free for two months from just an email address, via a no-code Poly Agent Builder or a real developer API (ADK with self-serve keys) — a real change from its previous sales-only-only access.
- A vendor-commissioned but methodologically disclosed Forrester Total Economic Impact study modeling 391% three-year ROI ($11.3M NPV, under-6-month payback) for a composite enterprise customer, including named, quantified benefit categories rather than one bare percentage.
Cons
- What happens after the 2-month self-serve free period is not publicly disclosed anywhere — no self-serve rate card, so budgeting past month two means either guessing or starting a sales conversation.
- PolyAI’s own reported GAAP financials show operating losses widening to $26.6M (from $23.8M) against $15M revenue for the 12 months to January 31, 2025 — spending well ahead of reported revenue even as forward ARR guidance (a less-audited metric) is projected near $40M.
- The Forrester ROI figures come from a PolyAI-commissioned study interviewing just 4 customers combined into one hypothetical composite organization — directionally useful, but not a broad independent sample, and its 5.0/5 G2 rating rests on a thin sample of only about 12 reviews.
Limitations
- The self-serve free tier is time-boxed to two months with no published rate for after that — real budgeting still requires either waiting to be billed or a sales conversation.
- Enterprise-scale deployments (the kind PolyAI is best proven at) still run through a separate, custom-quoted, sales-led engagement rather than the new self-serve on-ramp.
PolyAI pricing
- Self-serve (new, May 2026)Free / for the first 2 months
- Enterprise (custom)Custom / per minute, sales-led
See current pricing on poly.ai ↗Compare PolyAI alternatives →
PolyAI specs
Pricing
- Pricing model
- freemium
- Free tier
- ✓ Yes
Capabilities
- Model / LLM
- Proprietary (Raven), or GPT-5/Claude/Gemini since May 2026
- Interface
- Web
- Public API
- ✓ Yes
- Open source
- ✗ No
Deployment
- Deployment
- Cloud
PolyAI review
PolyAI is the pre-LLM-era incumbent of the voice-agent category — founded in 2017, years before Vapi, Retell AI, Bland AI or ElevenLabs Conversational AI existed — with real enterprise scale to show for it: 200+ customers, thousands of live deployments, a $750M valuation, and a Forrester study (vendor-commissioned, but methodologically disclosed) modeling 391% three-year ROI for a composite enterprise customer. Until May 2026 it was also the most closed platform in this category — sales-led only, no public price, no self-serve account. That changed: PolyAI now offers a genuine self-serve tier, free for two months from just an email address, with both a no-code builder and a real developer API (ADK). The catch is that PolyAI has not disclosed what happens once that free period ends, and its flagship enterprise deployments (like PG&E’s 16M-call-a-year operation) still run through a separate, custom-quoted sales engagement. Pick PolyAI if you want proven enterprise scale with a low-friction way to try it first; go in aware that the free tier’s long-run economics are still a question mark.
The pre-LLM-era enterprise incumbent that opened a real self-serve, free-to-start tier in May 2026 — but what happens after the free period is still undisclosed.
Frequently asked questions
Does PolyAI have a free trial or self-serve pricing?
Who founded PolyAI and when?
Does PolyAI have a developer API or SDK?
How is PolyAI different from Vapi, Retell AI, Bland AI, ElevenLabs Conversational AI or LiveKit Agents?
What proof does PolyAI publish that its product actually works?
Get a quote from PolyAI
Tell us what you need. We make the introduction to PolyAI.