## Summary so far
- Meeting purpose: Give participants a chance to share skills and practice pitching to different stakeholders within an account.
- Colin Heilbut (Rime) introduced himself as the newest salesperson at Rhyme with experience in Speech AI, including work at Deepgram and other AI labs.
- Participants introduced themselves and their roles at Assurant, including the Chief Product Officer, head of customer-facing voice, engineering lead, and contact center operations manager.
- Discussion touched on Assurant's focus on customer experience and primary business in warranty services.
- Assurant serves multiple clients including T-Mobile and Geico, with a core product group managing digital customer interactions.
- Contact center operations manage about 2,000 seats, balancing between tech and voice agents.
- Colin proposed an agenda covering introductions, Rhyme differentiation from other TTS solutions, deep dive on VoiceAI at Assurant focusing on voice quality issues, and a demo with Q&A.
- Colin uses a secure encrypted note-taking system during the meeting to capture pain points and success criteria.
- Rhyme's founding team includes members from Stanford with linguistic and engineering backgrounds.
- Rhyme differentiates by using thousands of hours of natural, full duplex conversations recorded with real people, rather than typical podcast or actor datasets.
- Rhyme handles tens of millions of calls monthly for large clients in restaurants, healthcare, telecom, and enterprise.
- A third-party study of 100,000 calls showed Rhyme has the lowest hang-up rate and fastest engagement compared to competitors, indicating superior customer interaction quality.
- Rhyme's technology has demonstrated case studies with 25% increase in call containment and 20% increase in sales for telecommunications clients.
- Assurant is in transition from traditional call centers to voice agents, currently using a smart IVR system.
- Participants seek more apples-to-apples comparisons relevant to their voice AI journey.
- Assurant currently uses NICE's Cotton Energy product for TTS but aims to develop fully conversational voice agents.
- Current voice agent experience at Assurant is somewhat deterministic and static; they are exploring more dynamic use cases.
- Rhyme's latest model, CODA, was developed from about 1,000 potential voices, categorized for client relevance, with a UI tool for voice style selection (formal, professional, American English, other languages).
- Assurant currently uses Microsoft-based TTS models via NICE Cotton Energy, with focus on pronunciation accuracy and customer complaints about voice quality.
- Rhyme emphasizes precise pronunciation and the ability to teach new words once for consistent replication across the system.
- Assurant balances 2,000 human contact center seats, aiming to transition more to voice agents while maintaining quality and control.
- Participants expressed caution about moving from static IVR trees to more agentic, dynamic voice AI solutions, emphasizing the need for control and reliability.
- Assurant aims to keep customers on AI interactions as long as possible to reduce human agent workload and improve capacity.
- Success metrics include maximizing information gathered before human handoff, task completion, and call containment.
- Some Assurant team members face layoffs, indicating organizational changes.
- Colin shared that Rhyme offers both cloud APIs in multiple regions and a fully on-premises deployment option managed via Docker containers, addressing compliance and infrastructure concerns.
- A live demo of Rhyme's voice agent was conducted, showcasing natural conversational flow, identity verification, claim submission, and language switching (English/French) with native speakers.
- Rhyme supports eight languages with native speakers for recordings and uses Deepgram STT to detect language and switch dynamically.
- Discussion on next steps focused on setting up a dev/test environment for trial, with a phased approach before any production deployment.
- Rhyme has native integration with NICE and can deploy on Microsoft Azure or other cloud/on-prem environments, supporting SOC 2 Type 2 compliance.
- Pilot success criteria to be scoped with Assurant, focusing on call completion, brand pronunciation, latency, and other relevant metrics.
- Decision timeline for Assurant is approximately three to four months for evaluation and decision-making.
- Colin emphasized Rhyme's commitment to high-quality AI-human experiences, distancing from poor AI interactions.
- Infrastructure discussion highlighted Assurant's compliance environment in Microsoft Azure data centers; Rhyme offers cloud APIs in multiple regions and fully managed on-premises Docker container deployments.
- Colin confirmed Rhyme's latency is among the fastest globally, balanced with high voice quality.
- Demo included a multi-turn conversation with identity verification, claim submission, and transfer to a live specialist.
- Language switching demo showed native French speaker voice and Deepgram STT language detection for dynamic switching.
- Next steps include setting up a constrained dev/test environment for Assurant to evaluate Rhyme voice AI.
- Security and approval processes for API and dev environment access discussed; baseline security checks and documentation required.
- Assurant team to help identify best voices and use cases for demo and pilot.
- Colin offered accessible customer service and direct contact availability for support.
- Discussion on prior Microsoft TTS evaluation process highlighted Rhyme's more hands-on, craftsman approach.
- Assurant team to lead pilot planning and timeline alignment; decision expected in 3-4 months.

## Key points
- The meeting is designed as a role-play to simulate pitching and conversation approaches with various stakeholders.
- Colin Heilbut has a background in GPU inferencing for speech recognition and legal audio processing.
- Participants emphasized the importance of voice quality as the first customer interaction.
- Assurant's primary business focus is on warranty services, with a strong emphasis on customer experience.
- Assurant manages digital interactions and contact center operations with a significant number of human agents.
- Concerns about voice quality at Assurant include latency, expressiveness, prosody, and customer complaints.
- Colin uses a secure encrypted note-taking system during the meeting to capture pain points and success criteria.
- Rhyme's founding team includes members from Stanford with linguistic and engineering backgrounds.
- Rhyme's approach is linguist-first, focusing on natural conversational data rather than standard datasets.
- Rhyme's technology shows measurable improvements in customer engagement and hang-up rates in large-scale third-party testing.
- Assurant is in the middle of transitioning from call centers to voice agents, currently using a smart IVR.
- Participants want more relevant, comparable data for their specific voice AI context.
- Data collection for Rhyme involves natural conversations, with categorization of voices for client relevance.
- Rhyme provides a UI tool to select voice styles (formal, professional, American English, other languages).
- Assurant currently uses NICE's Cotton Energy TTS product, based on Microsoft models, focusing on accuracy and pronunciation.
- Customer complaints about voice quality exist; Assurant seeks improvements while maintaining accuracy.
- Rhyme's system allows teaching new words once for precise, consistent pronunciation replication.
- Assurant's team is cautious about moving from static IVR trees to dynamic voice agents, emphasizing the need for control and manual testing.
- The transition to agentic voice AI solutions is recognized as inevitable but requires careful management.
- Assurant aims to maximize AI interaction time to reduce human agent workload and improve operational capacity.
- Success metrics include task completion, call containment, and minimizing customer hold or transfer times.
- Organizational changes at Assurant include potential layoffs affecting some team members.
- Compliance and infrastructure considerations are critical; Assurant operates in Microsoft Azure data centers and values options for deployment location.
- Rhyme supports compliant environments with cloud APIs in multiple regions and on-premises deployment options.
- Rhyme demo showed natural, multi-turn conversational capabilities including identity verification and claim processing.
- Language switching is handled by native speakers and STT language detection, supporting multiple languages including French.
- Pilot and testing phases are preferred before production rollout, with a typical decision timeline of 3-4 months.
- Rhyme offers native NICE integration and flexible deployment on cloud or on-premises with SOC 2 Type 2 compliance.
- Latency is among the fastest globally, balanced with high voice quality.
- Rhyme emphasizes a craftsman, customer-service oriented approach with accessible support.
- Security review and baseline checks are standard for pilot and API access.
- Prior Microsoft TTS evaluation was less hands-on compared to Rhyme's approach.

## Decisions
- Agenda approved by participants: introductions, Rhyme differentiation, VoiceAI discussion focusing on voice quality, demo, Q&A, and next steps.
- Next steps include setting up a dev/test environment for Rhyme voice AI trial.
- Pilot success criteria to be scoped collaboratively, focusing on call completion, brand pronunciation, latency, and other relevant metrics.
- Decision timeline set at approximately three to four months for Assurant to evaluate and decide on adoption.

## Action items
- Colin to continue discussing Rhyme's differentiation from other TTS solutions.
- Colin to lead a deep dive on VoiceAI applications at Assurant, focusing on voice quality issues.
- Participants to provide feedback and questions throughout the meeting.
- Colin to demonstrate Rhyme's capabilities and leave time for questions and next steps discussion.
- Colin to provide more detail later on Rhyme's voice data collection and voice selection process.
- Patrick to share more about Assurant's current TTS solution and voice agent use cases.
- Colin to explain Rhyme's approach to pronunciation accuracy and teaching new words in the system.
- Colin to address concerns about moving from static IVR to dynamic voice agents and discuss control mechanisms.
- Colin to clarify deployment options including cloud regions and on-premises solutions to address compliance and infrastructure needs.
- Patrick and Assurant team to explore approval process and security requirements for API and dev environment access.
- Colin and Rhyme team to assist in setting up a constrained demo/test environment for Assurant to evaluate voice AI.
- Assurant to identify key voices and use cases for initial testing and demo.
- Colin to coordinate with commercial and technical teams to support pilot rollout and integration with NICE.
- Patrick and team to lead next steps on pilot planning and timeline alignment.
- Colin to provide direct contact info post-meeting for ongoing support and communication.
- Assurant to prepare for security reviews and documentation requests to facilitate pilot approval.
- Colin to gather feedback on additional language support priorities from Assurant.

## Open questions
- Specific voice quality pain points at Assurant to be clarified during deep dive (e.g., latency, expressiveness, prosody).
- How to best compare Rhyme's conversational data to call center speech types for relevant benchmarking.
- Clarification on differences in speech style and requirements between natural conversations and call center interactions.
- Details on current limitations of NICE Cotton Energy TTS and what improvements Assurant seeks in fully conversational voice agents.
- How Rhyme's pronunciation teaching approach can address Assurant's accuracy and customer complaint concerns.
- How to ensure control and reliability as Assurant transitions from static IVR trees to more dynamic, agentic voice AI solutions.
- How to measure and improve success metrics such as task completion, call containment, and minimizing hold times.
- Impact of organizational changes, including layoffs, on voice AI adoption and team capacity.
- Further details on compliance requirements and deployment preferences for voice AI infrastructure.
- What languages beyond the current eight are priorities for Assurant to support in voice AI.
- What is the approval process and timeline for accessing APIs and deploying pilot environments within Assurant's compliance framework.