## Summary so far
- Colin Heilbut introduced himself as the newest account executive at Rhyme, with a background in speech recognition and AI, including early work at Deepgram and a French Voice AI Lab.
- Participants introduced themselves:
  - Maya runs contact center operations, focusing on call quality, handle time, and customer experience.
  - David works in IT and Risk, focusing on reliability, data handling, integration, and measured results before scaling.
  - Priya works in product for digital service, focusing on voice integration in web and app workflows and smooth handoffs.
- Colin uses a digital whiteboard to organize thoughts and ensure all concerns are addressed.
- Colin plans to briefly overview Rhyme's specialty but mainly focus on discovering key issues and concerns from participants.
- A short live demo is prepared for later in the meeting.
- Colin provided context on Rhyme's approach since 2022, emphasizing a unique method of creating hyperrealistic speech by recording in their own studio to produce high-fidelity, full duplex conversations.
- Rhyme powers tens of millions of calls monthly across hospitality, healthcare, telecom, and enterprise support sectors.
- A recent independent government survey showed Rhyme's TTS system had the lowest hang-up rate and fastest time to completion out of 100,000 outbound calls.
- Participants currently use a mix of vendors for speech generation but experience issues with flat, robotic-sounding calls that impact trust.
- Integration must be straightforward and supportable, not a "science project."
- Participants want voice experiences to feel connected and consistent across channels without jarring tone shifts.
- Flat, robotic calls are noticeable especially in longer calls, causing some customer impatience and impacting quality assurance metrics.
- Impact of poor voice quality includes increased call time due to customers talking over prompts or asking for repeats, reducing trust and increasing handle time.
- Success for Maya means improved trust and reduced call time; for David, better customer experience without added risk and clean integration; for Priya, a consistent, cost-effective customer journey.
- Success metrics include reduced handle time, steady CSAT and QA scores, positive agent feedback, and smoother conversations.
- Bottom line business impact focuses on lowering cost to serve and improving customer loyalty and trust.
- Customer trust is measured by CSAT, containment rates, fewer repeats or escalations, and task completion rates.
- Cost per interaction is measured by handle time, containment rates, and support overhead.
- David emphasized the need for clear success criteria, limited exposure during pilots, measurable outcomes, and a clear owner on their side to track metrics.
- Priya highlighted the importance of careful rollout to maintain consistency across customer journeys.
- Participants requested examples of measurable improvements after switching to Rhyme.
- Additional participant points:
  - Expressiveness is important; different call types (claims, device damage, warranty denials) require different tones.
  - Implementation speed is critical; going live in weeks rather than months is a strong advantage.
  - Brand voice customization is important to reflect company-specific tone.
- Colin shared an example of a large retail pizza client reporting a 20% increase in sales scale after switching to Rhyme.
- Colin conducted a live demo simulating a device warranty claim call, showcasing natural voice, smooth handoffs, empathy, and context memory.
- Participants gave positive feedback on the demo’s naturalness, flow, and personalization, suggesting minor improvements like a warmer opening and more cadence variation during fact collection.
- Colin explained Rhyme’s voice selection process based on user testing and use case matching.
- Discussion covered pilot structure: Maya sponsors, David leads technical assessment, Priya helps define use cases and success metrics.
- Typical pilot duration suggested is 4-6 weeks with measurement against current performance.
- If pilot metrics are positive, next steps include internal approvals (security, procurement, executives).
- Colin assured strong hands-on support during pilots via Slack, phone, and close integration with NICE orchestration.
- Participants identified latency as a key technical concern to discuss next.
- Participants confirmed pilot involvement would include contact center leadership, enterprise architecture, InfoSec, procurement, and executive sponsors.
- Success criteria prioritized as CSAT, handle time, containment/task completion, with reliability as a baseline requirement.
- Final go-live approval is a cross-functional decision with one executive owner informed by Maya’s recommendation and David’s technical sign-off.
- Colin invited further technical and business questions, focusing next on latency, voice selection, and benchmarking quality before production.
- Colin explained latency trade-offs: very low latency is critical but must balance with voice quality; Rhyme offers some of the lowest latency in the industry, especially with self-hosted local LLMs.
- Voice quality and responsiveness must be balanced to maintain user confidence.
- Voice selection is supported by a user interface and scientific user testing to match voices to use cases, geographies, and demographics.
- Benchmarking quality before production involves structured user studies.

## Key points
- Colin has extensive experience in speech recognition and AI.
- Maya's role centers on optimizing contact center performance and customer experience.
- David's role involves IT and Risk management, emphasizing reliability and data integration.
- Priya focuses on product development for digital services, ensuring smooth voice integration.
- Colin aims to tailor the discussion to participants' key concerns and use cases.
- Rhyme's differentiated approach involves proprietary studio recordings for more natural, high-fidelity conversational AI.
- Rhyme is actively deployed in multiple industries, handling millions of calls monthly.
- Independent study confirms Rhyme's TTS system reduces hang-ups and speeds call completion.
- Current pain points include robotic, flat-sounding calls that reduce customer trust.
- Integration complexity and inconsistent voice experiences across channels are key concerns.
- Poor voice quality leads to longer calls and lower customer satisfaction.
- Success criteria vary by role but focus on trust, efficiency, risk mitigation, integration, and consistent user experience.
- Key KPIs include handle time, CSAT, QA scores, containment rates, repeats, escalations, and agent feedback.
- Business impact centers on cost to serve, customer loyalty, and risk reduction.
- Pilots require clear success criteria, limited risk, measurable outcomes, and dedicated ownership.
- Rollouts must ensure consistent voice experience across customer journeys.
- Participants seek real-world validation beyond polished demos.
- Expressiveness and tone range are important for different call types.
- Fast implementation timelines and brand voice customization are strong priorities.
- Rhyme demonstrated a 20% sales scale increase for a retail pizza client.
- Live demo showed natural voice, smooth handoffs, empathy, and context memory.
- Pilot structure involves sponsorship, technical assessment, use case definition, and timeboxed measurement.
- Rhyme provides hands-on support during pilots and collaborates with NICE for orchestration integration.
- Latency is a key technical concern raised by participants.
- Pilot involvement includes contact center leadership, enterprise architecture, InfoSec, procurement, and executive sponsors.
- Success criteria prioritized as CSAT, handle time, containment/task completion; reliability is foundational.
- Final go-live approval is a cross-functional decision with executive ownership.
- Latency must be balanced with voice quality; Rhyme offers low latency, especially with self-hosted local LLMs.
- Voice selection is guided by a user interface and scientific user testing.
- Quality benchmarking before production involves user studies.

## Decisions
- None yet.

## Action items
- Colin to conduct a short live demo of Rhyme after discussion of key issues. (Completed)
- Colin to provide examples of measurable improvements after switching to Rhyme. (Provided example of 20% sales increase)
- Colin to continue gathering feedback on voice naturalness and pilot support details.
- Colin to address latency concerns and other technical questions in upcoming discussion.

## Open questions
- What is Rhyme’s approach to latency and other technical performance factors? (Partially answered: balance latency and quality; low latency with self-hosted local LLMs)
- How do clients typically choose voices? (Answered: user interface and scientific user testing guide voice selection by use case, geography, demographics)
- How is quality benchmarked before production? (Answered: structured user studies)
- Are there any additional teams involved in pilots? (Answered: contact center leadership, enterprise architecture, InfoSec, procurement, executive sponsors)
- Who makes the final go-live decision? (Answered: executive owner with cross-functional input)