PolyAI
PolyAI develops AI voice agents for large enterprises, designed to handle real phone conversations across customer service, call routing, bookings, payments, utility outages and regulated service workflows.
Europe's most interesting startup stories. Every week.
No spam. Unsubscribe anytime.
Podcast Episode
🎙️ Deep Dive
EP58 — Unicorn Files — PolyAI: the European voice AI company rewriting the call center
Listen to the episodeEP49 — Unicorn Files — Entrepreneurs First: talent before the idea
Listen to the episodeAbout PolyAI
PolyAI is a British enterprise voice AI startup founded in London in 2017 by Nikola Mrkšić, Pei-Hao “Eddy” Su and Tsung-Hsien “Shawn” Wen, machine learning researchers connected to the University of Cambridge ecosystem. The company builds voice agents for large enterprises handling high call volumes and complex operational conversations. Its platform is not a chatbot adapted to voice, but a dialog-first system: low latency, accent and interruption handling, human fallback, enterprise integrations, auditability and compliance. Its core product is the Agentic Dialog Platform, supported by Raven, a proprietary model trained on more than one billion enterprise conversations. Public customers include Marriott, Caesars Entertainment, PG&E, UniCredit, Foot Locker and FedEx. In December 2025, PolyAI announced an $86 million Series D, bringing total funding above $200 million. The $750 million valuation was reported by SiliconANGLE, but was not disclosed in PolyAI's official announcement.
The Story
PolyAI emerged from the founders’ work in dialog systems and machine learning at the University of Cambridge. The original insight was that enterprises should not have to choose between customer service scalability and conversation quality: the phone channel remains critical in sectors such as hospitality, utilities, banking and healthcare, while traditional IVR systems fail to handle natural interaction, context and operational complexity.
How PolyAI works
Business Model
B2B enterprise SaaS/deployment model: selling to large companies through measurable proof-of-value projects, implementations on high-volume use cases, integrations with contact-centre systems and land-and-expand across flows, languages, countries and business units.
Revenue Model
Revenue from enterprise contracts, SaaS platform access, deployment, usage volumes and implementation/governance services.
Products & Services
Key Products
- Agentic Dialog Platform
- Raven
- Agent Builder
- Agent Development Kit
- Agent Studio
- Voice AI Agents
Core Use Cases
- Inbound customer service calls
- Natural language call routing
- Booking and reservations
- Billing and payments
- Account management
- Authentication
- Utilities outages
- FAQ automation
- Troubleshooting
- Customer data and interaction insights
Market & Clients
Key Customers
Serves enterprise customers such as Marriott, Caesars Entertainment, PG&E and UniCredit/Zagrebačka banka (plus Foot Locker, FedEx, Fogo de Chão, Golden Nugget and Simplyhealth) to automate high-volume voice conversations in contact centers.
Geographic Presence
How PolyAI competes
Competitors
Competitive Advantages
- Dialog-first architecture rather than chat-first voice wrapper
- Proprietary Raven model trained on large-scale enterprise conversations
- Enterprise customer references in regulated and high-volume environments
- Integration with existing contact-center and enterprise systems
- Governance, fallback and auditability positioned as part of the platform
- Proof-of-value sales motion based on containment, CSAT, hours saved and ROI
How PolyAI grows
Growth Strategy
Enterprise land-and-expand growth, international expansion, opening the platform to builders and developers, strengthening go-to-market and developing the Agentic Dialog Platform/Agent Studio.
Distribution Model
Enterprise sales, proof of value, customer case studies, partner/integration motion and developer/platform expansion.
Moat (Defensibility)
Combination of enterprise conversational data, the proprietary Raven model, regulated production use cases, integrations, enterprise buyer trust and continuous learning from deployments.
Regulatory Context
Rilevanti GDPR, AI Act, DORA per financial services, PCI DSS per pagamenti, HIPAA per sanità USA, SOC 2 e requisiti enterprise di sicurezza, audit, data retention, escalation e vendor risk management.
Key Risks
- Commoditizzazione dei modelli vocali e degli LLM multimodali
- Competizione di Big Tech e incumbent del contact center
- Compressione dei margini se il prodotto richiede delivery troppo intensivo
- Rischi reputazionali da errori conversazionali in contesti sensibili
- Vendor lock-in percepito dai buyer enterprise
- Obblighi di trasparenza, audit, privacy e resilienza nei verticali regolati
- Differenza tra metriche di ROI dei clienti e disclosure finanziarie ufficiali
Strategic Insights
- The potential moat is not only the voice model, but the combination of enterprise conversational data, regulated deployments, integrations, buyer trust and continuous call-flow improvement.
- The Series D with NVentures strengthens PolyAI's credibility in a market where enterprise buyers look for signals of vendor scalability and durability.
- The platform shift through Agentic Dialog Platform, Agent Builder and ADK suggests a move from managed deployments to a more scalable product for CX, operations, product and developer teams.
- The reported $750 million valuation places PolyAI below unicorn status, but with a clear European rising-star profile in enterprise AI agents.
- The main risk is competitive compression: Big Tech, contact-centre incumbents and agentic AI startups could erode the advantage if Raven, data and deployment capability do not become defensible.
Funding & Investors
Total Funding
Latest Valuation
Funding Rounds
- 2025 - Series D: $86M
Lead: Georgian, Hedosophia, Khosla Ventures
Key Investors
Founding Team
Founders
Key Executives
- Nikola Mrkšić - CEO
- Tsung-Hsien “Shawn” Wen - CTO
- Pei-Hao “Eddy” Su - SVP Engineering
Key Metrics
Lessons from PolyAI
- Start from a painful and measurable workflow, not from a technology demo.
- Sell operational ROI before AI magic: containment, CSAT, abandonment rate, labour hours saved and payback.
- Build governance, audit logs, data retention, policy and human fallback before scaling into regulated verticals.
- Use enterprise case studies as a commercial asset: PG&E and UniCredit are stronger than generic demos.
- In enterprise B2B, product and delivery must scale together: the risk is not only technical, but organisational.
Similar Startups
Fresha
London-based platform for salons, spas and wellness operators managing bookings, payments, marketplace and business operations.
Read AnalysisConduct
London startup founded by ex-Palantir engineers building an AI OS to understand, operate and modify systems like SAP, Salesforce and Oracle.
Read AnalysisFrontier Health
London healthtech developing Juno, an AI agent supporting hospital administrative staff.
Read AnalysisSources
- PolyAI — Official website — PolyAI
- PolyAI — About — PolyAI
- PolyAI raises $86M to transform how enterprises talk to their customers — PR Newswire / PolyAI
- Call center chatbot startup PolyAI raises $86M at a $750M valuation — SiliconANGLE
- PolyAI opens its Agentic Dialog Platform — PR Newswire / PolyAI
- How Pacific Gas and Electric saved 35,000 labor hours with PolyAI — PolyAI
- Improving NPS by 14 points with UniCredit — PolyAI
- AI Act — European Commission
- Digital Operational Resilience Act (DORA) — EIOPA