## Summary so far
- Meeting began with introductions.
- Participants include Monique (Operations), Raj (Contact Center Tech Stack), Elena (CX and Insights), and Colin (new account executive at Rhyme with speech AI experience).
- Colin plans to present how Rhyme works, what makes it special, and focus on the participants' needs.
- Participants emphasized the need for practical, concrete solutions over flashy features due to current pressures and ongoing issues.
- Integration effort is a key concern given the existing, opinionated tech stack.
- Rhyme is compatible with Cognizant and can be integrated with minimal heavy lifting, mostly involving UI toggles and security checks.
- Colin will use a Copilot system for secure, local encrypted transcription during the meeting to capture notes accurately.
- Participants are cautious about recordings; Colin clarified no recordings are stored or used beyond the meeting.
- Rhyme was founded in 2022 by a team including linguists and engineers from Stanford and Harvard.
- Rhyme differentiates itself through custom-built data collection using a recording studio with real people, rather than common podcast or actor datasets.
- Rhyme processes tens of millions of calls monthly, mostly inbound contact center calls.
- Recent study of 100,000 calls showed Rhyme had the lowest hang-up rate and fastest call completion time compared to competitors.
- Rhyme helps customers improve sales and performance metrics.
- Current speech AI tools in use include a mix of Nice on the Court and other vendors; voice quality is okay but can feel robotic and flat, impacting caller trust.
- Trust is critical in claims context; robotic voice leads to hang-ups or requests for human agents, increasing costs and handle time.
- Ease of rollout is a priority: short implementation, low engineering lift, minimal risk of downtime or escalation spikes.
- Success criteria for voice AI include higher containment without hurting CSAT, fewer negative voice comments, no increase in handle time.
- Operations success criteria include reducing repeat calls, avoiding new edge cases, no new fires, and proving effectiveness with small pilots.
- Technical success criteria include reliability, clean integration, and compatibility with existing systems like Nice.
- Current voice AI setup includes Nice on the Court and other vendors; quality issues include robotic, flat voice lacking empathy.
- Participants want measured improvement before considering change; not interested in change for its own sake.
- Risk and rollout effort are top concerns; small pilots with clear metrics preferred before scaling.
- Demo started showcasing Rhyme’s conversational AI handling a claims scenario with natural, empathetic dialogue.
- Demo showed Rhyme’s AI guiding a user through a device damage claim, verifying details naturally, confirming information, and smoothly transferring to a human specialist with shared context.
- Demo call was smooth, natural, with steady pacing and clear confirmation steps that build trust.
- Participants suggested possibly trimming some confirmation lines for efficiency but overall found the demo solid.
- Voice quality in demo was noted as better than current solutions.
- Rhyme’s latest model, Coda, was trained with thousands of voices and refined through rigorous pairwise comparisons to select about 100 high-quality voices.
- Desired voice qualities for the brand: warm but competent, not overly cheerful or call center-like, reassuring and professional.
- Decision process for pilot success involves operations, IT, CX, and procurement; consensus-driven with risk as a primary concern.
- Additional teams involved: IT security and possibly legal for data handling.
- Early planning should include security review, documentation on data handling and encryption.
- Procurement is not a blocker if business case is strong; security review may take several weeks.
- Reference checks with similar customers expected.
- Pilot timeline estimated at 4-8 weeks if security review proceeds smoothly.
- Monique prefers a narrow scope pilot first with data-driven go/no-go decision.
- Final sign-off requires consensus from Raj (technical/security), Elena (CX), Monique (operations), procurement, and an executive sponsor.

## Key points
- Monique runs operations and may need to leave intermittently due to a busy day.
- Raj is responsible for the contact center technology stack, prioritizing uptime and roadmap stability.
- Elena focuses on customer experience, quality, and impact on metrics.
- Colin has nearly four years of experience in speech AI, previously at Deepgram, Intelligent Voice, and Gradium.
- The team is under pressure about voice quality but is dealing with many urgent issues ("fires").
- Integration effort is critical due to an already opinionated tech stack.
- Rhyme is compatible with Cognizant and can be integrated with minimal heavy lifting, mostly involving UI toggles and security checks.
- Colin will use a Copilot system for note-taking during the meeting.
- Participants prefer concrete, practical solutions rather than flashy features.
- Rhyme’s unique approach to data collection involves custom recordings in a studio setting, aiming for higher voice quality.
- Rhyme serves large clients across industries including restaurants, healthcare, telecommunications, and governments.
- Rhyme’s technology is proven by large-scale deployments and performance studies.
- Current speech AI tools feel robotic and lack empathy, causing trust issues and increased costs.
- Ease of rollout is measured by short implementation time and low engineering lift, with minimal risk of downtime or escalation spikes.
- Success metrics include containment rates, CSAT, handle time, repeat calls, escalation rates, and supervisor escalations.
- Reliability and clean integration with existing systems like Nice are essential.
- Current providers include bundled platform TTS and one other vendor for specific flows.
- Participants want to avoid new edge cases, fires, or escalation spikes.
- Small pilots with measurable success are preferred before wider rollout.
- Demo illustrates Rhyme’s conversational AI handling claims with natural language, verifying details empathetically, and transferring to human specialists with full context.
- Demo call was smooth and natural with steady pacing and clear confirmation steps.
- Participants noted voice quality in demo was better than current solutions.
- Rhyme’s latest model, Coda, was trained with thousands of voices and refined to about 100 high-quality voices through pairwise comparison.
- Desired voice qualities: warm, competent, reassuring, professional, not overly cheerful or call center-like.
- Decision process for pilot success is consensus-driven involving operations, IT, CX, and procurement, with risk as a key factor.
- IT security and legal teams may need to be involved for security and data handling reviews.
- Security review and documentation on data handling and encryption should be planned early.
- Procurement is not a blocker if the business case is strong; security review may take several weeks.
- Reference checks with similar customers are expected.
- Pilot timeline estimated at 4-8 weeks if security review proceeds smoothly.
- Monique prefers a narrow scope pilot first with data-driven go/no-go decision.
- Final sign-off requires consensus from Raj (technical/security), Elena (CX), Monique (operations), procurement, and an executive sponsor.

## Decisions
- Pilot to be small and narrowly scoped with clear success metrics.
- Consensus-driven decision process involving operations, IT, CX, procurement, and executive sponsor.
- Early engagement with IT security and possibly legal for data handling and encryption documentation.
- Procurement not a blocker if business case is strong.
- Reference checks with similar customers will be conducted.

## Action items
- Colin to provide an edited summary of meeting notes for participant review and confirmation.
- Colin to prepare security documentation and data handling details for IT security review.
- Colin to plan for early engagement with IT security and legal teams regarding security and data handling.
- Colin to prepare a clear pilot plan with success metrics as requested by Monique.
- Colin to arrange reference checks with similar customers to support procurement and security processes.
- Colin to send a short, high-level summary of pilot success criteria for internal participant review.

## Open questions
- Are there any additional procurement or legal requirements beyond security review and data handling documentation?
- What is the expected timeline for security review and procurement approval from the participants' side?
- Are there specific success metrics or KPIs beyond those discussed that the team wants to track during the pilot?