Avantive Deploys AI Voice Tools as Call Agents Prefer Machine Feedback
Ai4
Standardized generative AI coaching and real-time assistance eliminate supervisor bias while driving full call-monitoring compliance across heavily regulated contact centers.
## Background
Published on YouTube by Ai4 on August 16, 2026, this session features RingCentral's Amir Hameed and Avantive Solutions CEO Frank Pettinato discussing how enterprises can move beyond basic automation toward orchestrating agentic voice AI across customer engagement channels. The presentation focuses on real-world production deployments, architectural decisions, governance models, and practical frameworks to overcome stalled pilots.
## TL;DR
Amir Hameed and Frank Pettinato outline how integrating voice AI across upfront automation, real-time agent assist, and post-call analytics drives tangible customer experience outcomes. Avantive Solutions details its transformation from legacy on-premise hardware to an AI-first BPO model, using automated quality scoring and generative role-playing bots to achieve a 174% increase in QA monitoring. The discussion highlights a significant cultural shift where contact center agents actively prefer objective AI coaching over supervisor critique, alongside private equity governance models that align executive KPIs with AI activation. Ultimately, real-time agent assist delivers rapid ROI by enforcing verbatim compliance and accelerating onboarding ramp times in highly regulated environments.
## Atmosphere & style
The session has a polished, pragmatic, and collaborative keynote dynamic, seamlessly blending strategic enterprise framing with frontline operational reality. Amir and Frank maintain an engaging and candid rapport before a massive audience, grounding transformative AI capabilities in concrete implementation lessons rather than speculative hype. The discussion feels authoritative and forward-looking, conveying a clear sense of operational maturity as AI transitions from experimental point solutions into core enterprise infrastructure.
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) Enterprise AI adoption has moved beyond theoretical debate to focus on tangible business outcomes, with voice remaining the primary and most human customer communication channel.
- [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) Enterprise AI initiatives frequently stall after pilot phases due to tactical point-solution sprawl and disconnected data layers that drop customer context during handoffs.
- [03:02](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=182s) An effective customer experience architecture integrates three AI phases: upfront self-service automation, context-rich handoffs to live agents, and post-interaction conversational intelligence.
- [04:57](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=297s) Strict regulatory compliance prohibits autonomous synthetic voice AI on outbound marketing calls, mandating that AI be directed toward assisting rather than replacing human representatives.
- [04:57](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=297s) Decommissioning legacy on-premise hardware and moving to cloud-native platforms is an essential operational foundation for adopting continuous AI updates and innovations.
- [07:49](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=469s) Avantive differentiates algorithmic functions by applying machine learning for predictive dialing analytics and reserving generative AI for live contextual workflows and agent assistance.
- [07:49](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=469s) Moving away from isolated point solutions requires an enterprise-wide transformation, repositioning the company at the board and executive levels as an AI-first organization.
- [10:45](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=645s) AI-powered voice analytics achieved a 174% increase in quality monitoring by expanding evaluation coverage from tiny manual spot-checks to virtually 100% of customer interactions.
- [10:45](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=645s) Automated quality scoring paired with agentic role-playing bots provides non-judgmental, standardized coaching and interactive practice tailored to specific agent weaknesses.
- [13:22](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=802s) Contact center culture shifted over 36 months from initial pushback to agents actively preferring objective, non-threatening AI feedback over traditional managerial critiques.
- [16:06](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=966s) Rapid 12-to-24-month AI development cycles create operational and valuation friction against traditional 3-to-5-year private equity investment holding timelines.
- [18:10](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=1090s) Real-time agent assist delivers higher immediate enterprise ROI than full autonomy by enforcing verbatim regulatory script compliance and drastically accelerating new-hire onboarding.
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) Opening Remarks and the Imperative for Agentic Voice AI
Amir Hameed kicks off the session before an audience of 12,000 attendees at the conference, playfully empathizing with anyone who got stuck waiting in long entry lines the previous day. Setting the stage for his discussion with Frank Pettinato, CEO of Avantive Solutions, Amir establishes that enterprise adoption of artificial intelligence is no longer a speculative debate—it is an immediate reality where the central question has shifted from "if" to "when, how much, and whether it actually produces business outcomes." While much of the industry focuses on text and digital automation, RingCentral is zeroing in on agentic voice AI. Amir emphasizes that voice remains the most natural and fundamentally human communication channel for engaging colleagues, partners, and customers.
Amir addresses why so many enterprise AI deployments hit a wall or stall out after initial pilot phases. Organizations frequently fall victim to tactical tool sprawl, buying isolated software solutions that fail to work together. This creates disjointed customer experiences where context is dropped during handoffs between automated bots and human representatives, forcing callers to re-identify themselves and repeat their issues. Furthermore, disconnected data layers between front-office interactions and back-office systems prevent critical customer insights from circulating effectively. Because "data is the fuel for AI," feeding poor or fragmented data into AI systems only amplifies operational friction and degrades the overall customer experience.
[03:02](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=182s) Orchestrating the End-to-End Customer Journey with AI
RingCentral conceptualizes customer experience not as a collection of fragmented touchpoints, but as an integrated, AI-orchestrated journey spanning three core phases: upfront automation, real-time agent empowerment, and post-interaction intelligence. Upfront, businesses deploy 24/7 self-service tools like AI receptionists and digital representatives to resolve inquiries instantly. When an issue requires human judgment or empathy, the system executes a seamless handoff to human agents via real-time Agent Assist. This ensures that every piece of background and conversational context gathered by the AI agent transfers instantly to the human representative, eliminating wasted time—especially for high-value or VIP customers.
subgraph FrontEnd["1. Upfront Automation"]
A["Inbound / Outbound Contact"] --> B["AI Receptionist & Digital Reps"]
end
subgraph Escalation["2. In-Flight Collaboration"]
B -- "Complex Query / Escalation" --> C["Seamless Context Transfer"]
C --> D["Human Agent + Agent Assist"]
end
subgraph BackEnd["3. Post-Interaction Intelligence"]
D --> E["Conversational Intelligence Engine"]
E --> F["100% Interaction Analysis & Insights"]
end
```
The final phase of this lifecycle leverages conversational intelligence after the call concludes. Rather than relying on legacy quality assurance practices that manually review a tiny sample of recorded calls, post-interaction AI inspects every single customer conversation at scale. This comprehensive visibility allows organizations to extract actionable operational insights, verify compliance, and continually refine the customer experience.
- **Upfront AI Agents:** AI receptionists and representatives handle routine tier-one interactions around the clock.
- **Context-Rich Handoffs:** Agent Assist equips representatives with historical data and digital session transcripts to avoid customer repetition.
- **Total Conversation Visibility:** Conversational intelligence analyzes 100% of voice and digital interactions across the enterprise.
[04:57](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=297s) Avantive Solutions and the Breaking Point with Legacy Infrastructure
Frank Pettinato introduces Avantive Solutions as a global business process outsourcing (BPO) firm with specialized focus across two core industry verticals: healthcare/life sciences and telecommunications. In telecommunications, Avantive conducts large-scale outbound telemarketing—a data-intensive and high-friction operational environment. Frank notes that strict regulatory compliance rules prohibit the use of autonomous synthetic AI voices on outbound marketing calls; consequently, Avantive directs its AI capabilities toward empowering and assisting human agents during live conversations rather than attempting to replace them.
Reflecting on Avantive’s 30-plus-year operating history and its decade-long relationship with RingCentral, Frank recounts the breaking point that forced the company away from its legacy on-premise Avaya infrastructure. Facing costly, mandatory hardware refresh cycles, Avantive decided to completely exit the "big iron data center business." The organization realized that self-funding and maintaining premise-based contact center hardware was an unsustainable long-term model that throttled innovation. Migrating to a cloud-native architecture enabled Avantive to eliminate infrastructure maintenance burdens and align with continuous software release cycles, setting the technological foundation necessary for advanced AI adoption.
[07:49](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=469s) Moving from Tactical Machine Learning to Strategic AI Transformation
Frank traces Avantive's algorithmic journey back five years, establishing a clear operational distinction between machine learning and generative artificial intelligence. Avantive utilizes machine learning engines primarily for predictive modeling and dialing analytics—evaluating real-time data minute by minute to optimize live contact rates and ensure agents reach the right prospects with the right offers. In contrast, generative AI platforms are reserved for dynamic content generation, contextual assistance, and taking agentic actions during live customer engagements.
Over the past three years, Avantive initially adopted AI through tactical point solutions, purchasing distinct tools to solve isolated problems such as quality management scoring, real-time agent guidance, or employee onboarding. However, entering 2026, the company completely overhauled its operating philosophy. Moving away from disjointed point solutions, Avantive elevated machine learning and AI into core enterprise pillars, repositioning itself from the board level to the Executive Leadership Team (ELT) as "an AI company that happens to be a call center." Executive discussions now balance the governance risks introduced by AI against the productivity gains it unlocks, maintaining an intense focus on consumer protection and regulatory adherence.
- **Machine Learning vs. GenAI:** Machine learning drives predictive outbound dialing strategies; generative AI powers real-time content and contextual agent workflows.
- **Dialing Strategy Optimization:** Algorithmic adjustments occur hour by hour and minute by minute to counter dropping live contact rates.
- **Evolution of AI Strategy:** Transitioned from buying standalone tactical point tools (2021–2024) to executing an enterprise-wide AI-first strategic transformation by 2026.
[10:45](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=645s) AI-Driven Quality Management and Interactive Agentic Training
Addressing regulatory dynamics, Frank Pettinato notes that the Federal Communications Commission (FCC) is swinging toward stricter restrictions, requiring organizations to navigate call center AI implementation carefully while keeping consumer service at the core. When Amir Hameed brings up Avantive Solutions achieving a 174% increase in quality monitoring, Pettinato explains the transformation across their thousands of contact center agents. Historically, supervisors could only evaluate a small sample of calls. Today, Avantive routes virtually every customer interaction through a voice analytics platform powered by AI to enrich conversation data and generate automated quality scorecards.
Initially, these automated scorecards were handed directly to team managers to conduct one-on-one reviews. However, Avantive discovered that human-to-human feedback often introduced stylistic friction and interpersonal tension. To solve this, they implemented a generative AI coaching system that delivers constructive, empathetic, and objective evaluations directly to agents in a standardized format. Rather than stopping at passive feedback, the platform automatically generates targeted training modules and assigns them to an agentic AI interactive bot. This bot role-plays customer scenarios tailored precisely to the agent's coaching needs, enabling self-paced, pressure-free practice.
```mermaid
flowchart TD
accTitle: AI-Powered Quality Assurance and Agentic Training Workflow
subgraph INGEST ["1. Interaction & Analysis"]
A["Customer Voice Call"] --> B["Voice Analytics Engine"]
B --> C["AI Quality Scorecard"]
end
subgraph COACH ["2. Feedback & Simulation"]
C --> D["Generative AI Coaching Feedback"]
D --> E["Personalized Training Modules"]
E --> F["Agentic Interactive Role-Play Bot"]
end
subgraph REVIEW ["3. Human Oversight"]
D -.-> G["Supervisor 1-on-1 Mentoring"]
end
```
- **174% Increase in Quality Monitoring:** AI analysis scales QA from selective spot-checking to virtually 100% call coverage.
- **Tailored Role-Play Simulations:** The agentic bot simulates realistic customer interactions specifically designed to remediate weaknesses flagged in the QA scorecard.
---
[13:22](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=802s) Overcoming Agent Resistance and the Cultural Shift Toward Machine Feedback
Transitioning from traditional oversight to AI-driven evaluation required navigating substantial workforce apprehension. Pettinato reflects on the dramatic cultural evolution over a 36-month span: when Avantive first rolled out agent-based AI tools three years earlier, they encountered significant disruption and employee pushback because AI was not yet mainstream. Fast forward to the present, and call center agents have grown remarkably comfortable with the technology, actively preferring automated machine feedback over direct human critique.
When asked whether this change was driven by employee turnover and a younger demographic or existing staff adapting, Pettinato emphasizes that the shift is predominantly cultural. Agents appreciate that generative AI delivers feedback in a warm, friendly, and non-judgmental tone, removing the anxiety and perceived bias of supervisor evaluations. While human coaches remain firmly in the loop for high-level guidance and mentorship, the daily cadence of machine evaluation aligns with personal self-improvement habits. Just as individuals routinely consult consumer AI tools for guidance on their health or personal lives to "become the best version of themselves," agents leverage enterprise AI feedback to excel at their craft.
- **36-Month Perception Shift:** Contact center staff evolved from active resistance to openly preferring AI-driven feedback over human managerial critiques.
- **Human-in-the-Loop Safeguards:** AI handles daily objective scoring and standardized feedback, while human leadership provides strategic 1-on-1 support and coaching.
---
[16:06](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=966s) Private Equity Governance and Enterprise-Wide AI Implementation
Transforming Avantive into an AI-first organization has altered board governance and executive accountability. As a private equity-backed enterprise, Avantive’s board meetings now open with dedicated discussions assessing AI market disruptions alongside potential operational gains in productivity, production capacity, and cost optimization. Pettinato highlights a growing tension in the investment landscape: private equity operates on three-to-five-year holding cycles, yet AI capabilities are shifting every 12 to 24 months, forcing investors and operators to constantly re-evaluate risk profiles for traditional business models.
To operationalize this strategy across the entire enterprise, members of the Executive Leadership Team (ELT) are measured against explicit key performance indicators (KPIs) tied to AI activation. Rather than restricting AI to customer-facing channels, Avantive actively targets repetitive, mundane processes across human resources information systems (HRIS), recruiting, and corporate accounting. Executive leadership reinforces transparent internal messaging to ensure employees understand that AI tools are designed to amplify human capability and drive corporate performance rather than eliminate jobs.
- **PE Investment Window Compression:** Technological evolution occurring every 12 to 24 months creates strategic friction against typical 3- to 5-year private equity investment cycles.
- **Executive AI KPIs:** Leadership incentives are directly linked to deploying AI automation across both core operations and back-office functions.
---
[18:10](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=1090s) Real-Time Agent Assist and Compliance Enforcement
Looking at the immediate horizon of agentic voice technology, Pettinato identifies AI agent assist as delivering the highest near-term ROI, outpacing fully autonomous agents in complex enterprise scenarios. Real-time guidance accelerates "speed to green" for newly onboarded agents while elevating the efficiency and throughput of tenured staff. In Avantive’s highly regulated operating environments, strict compliance requires agents to recite precise legal terminology in mandatory sequences. Real-time agent assist ensures verbatim script compliance while active calls are underway.
Avantive leverages RingCentral's AI Conversational Expert (powered by RingSense) to provide agents with instant post-call summaries and performance insights the moment a conversation ends. This instantaneous loop feeds seamlessly into the broader quality management ecosystem, generating rich feedback and immediate action plans. Hameed closes by inviting attendees to visit RingCentral at Booth 1025 to experience live demonstrations of agentic voice tools, build interactive AI agents, and explore production-ready agent assist capabilities.
- **Speed to Green:** Real-time AI guidance shortens onboarding ramps and accelerates time-to-proficiency for new call center hires.
- **Strict Verbatim Compliance:** Real-time speech monitoring validates mandatory legal disclosures and conversational sequence requirements.
- **RingSense AI Conversational Expert:** Generates immediate post-interaction telemetry and feeds data into comprehensive quality action plans.
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) when, how much, and whether it actually produces business outcomes. — Amir Hameed
> [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) data is the fuel for AI — Amir Hameed
> [04:57](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=297s) big iron data center business — Frank Pettinato
> [07:49](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=469s) an AI company that happens to be a call center. — Frank Pettinato
> [13:22](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=802s) become the best version of themselves — Frank Pettinato
> [18:10](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=1090s) speed to green — Frank Pettinato
## Mentioned
Amir Hameed, Frank Pettinato, RingCentral, Avantive Solutions, Ai4, Avaya, Federal Communications Commission (FCC), RingSense, AI Conversational Expert, Agent Assist
## Video description
Amir Hameed - Senior Vice President - Global Technical Solutions, RingCentral
Frank Pettinato - CEO, Avantive Solutions
Most enterprises have automated customer interactions, yet few have orchestrated them. Agentic Voice AI goes beyond automation, reasoning across conversations, enterprise data, workflows, and human expertise to achieve outcomes. This session explores how enterprises move from AI agents that respond to requests to agentic systems that orchestrate intelligent customer engagement across voice and digital channels. Drawing on real-world deployments, we'll examine the architectural decisions, governance, and operational practices that distinguish successful production deployments from stalled pilots. Attendees will leave with actionable frameworks and scaling considerations you can apply immediately.
Amir Hameed of RingCentral and Frank Pettinato of Avantive Solutions argue that enterprise customer experience requires replacing disconnected point tools with an integrated, agentic voice strategy. By combining predictive machine learning, real-time agent guidance, and automated quality coaching, companies can meet strict regulatory requirements while scaling contact center productivity.
Orchestrating the End-to-End Voice Journey (3:02)
Amir Hameed observes that enterprise artificial intelligence pilots frequently stall because organizations buy isolated point tools. This software sprawl leads to disconnected data layers and dropped context during transfers between automated bots and human representatives, forcing callers to repeat information. Because data fuels AI, fragmented data ultimately amplifies operational friction.
To address this, RingCentral structures customer experience across three connected phases: upfront 24/7 automation with AI receptionists, context-rich escalations to human agents equipped with real-time Agent Assist, and post-interaction conversational intelligence. Rather than relying on legacy quality assurance practices that manually inspect minimal call samples, the post-call engine analyzes 100 percent of voice and digital interactions across the enterprise to verify compliance and extract operational trends.
Moving Beyond Hardware to Predictive Intelligence (4:57)
Operating across healthcare and telecommunications, Avantive Solutions reached a breaking point with its legacy on-premise Avaya infrastructure. Facing expensive, mandatory hardware refresh cycles, the firm exited premise-based data centers to adopt a cloud-native platform with continuous updates, creating the technological foundation required for advanced AI.
Avantive draws an operational distinction between machine learning and generative artificial intelligence. The firm utilizes machine learning for predictive dialing analytics, making minute-by-minute algorithmic adjustments to optimize live contact rates. In contrast, generative AI powers live contextual guidance and agent workflows. Between 2021 and 2026, Avantive moved away from tactical point tools to reposition itself at the board level as an AI-first organization balancing productivity gains against governance risks.
Automated Quality Scoring and Agentic Coaching (10:45)
Because Federal Communications Commission regulations restrict synthetic AI voices on outbound telemarketing calls, Avantive concentrates AI capabilities on assisting human agents. Pettinato highlights a 174 percent increase in quality monitoring achieved by routing interactions through voice analytics that generate automated scorecards across virtually all calls.
To resolve interpersonal friction caused by supervisor critiques, Avantive introduced generative AI coaching that delivers objective evaluations directly to representatives in a standardized format. The system then automatically assigns personalized training modules to an agentic role-play bot. This interactive bot simulates realistic customer conversations tailored to the agent's identified weaknesses, providing self-paced, pressure-free practice while reserving managers for high-level mentorship.
Cultural Shift, Board Governance, and Compliance (13:22)
Workforce attitudes transformed substantially across a 36-month timeline: agents shifted from resisting early AI deployments to actively preferring machine evaluations over managerial critiques due to the non-judgmental tone. Organization-wide, Avantive measures its Executive Leadership Team against explicit key performance indicators tied to deploying AI across operational channels as well as internal back-office functions like HRIS, recruiting, and accounting.
Pettinato notes strategic friction in private equity governance, where 12-to-24-month technology development cycles challenge traditional three-to-five-year investment horizons. In the near term, Pettinato claims real-time agent assist delivers higher immediate returns than fully autonomous voice agents by enforcing verbatim legal script compliance on active calls, accelerating onboarding ramps, and generating instant post-call summaries via RingSense conversational tools.
While real-time agent assist offers demonstrable ROI and compliance safeguards in complex call center environments, operators must continually adapt to evolving regulatory restrictions and a rapid 12-to-24-month AI cycle that compresses traditional corporate holding strategies.
## Background
本内容由 Ai4 于 2026 年 8 月 16 日发布在 YouTube 平台上。视频记录了 RingCentral 全球技术解决方案高级副总裁 Amir Hameed 与 Avantive Solutions 首席执行官 Frank Pettinato 的炉边对话，探讨企业如何超越传统自动化，利用 Agentic Voice AI 协同对话、数据、工作流与人工专业知识以重塑客户体验。
## TL;DR
对话深入探讨了 BPO 巨头 Avantive Solutions 如何告别传统本地硬件架构，并与 RingCentral 深度合作将 AI 升级为核心企业战略。双方详细拆解了机器学习在拨号预测上的应用，以及生成式 AI 在质检覆盖率提升 174%、自动化教练反馈与智能体模拟实操中的闭环实践。此外，嘉宾还分享了如何克服员工对 AI 监控的抵触情绪、在 3-5 年私募股权周期内推进 AI 治理，并借助实时座席辅助（AI Assist）在严苛合规场景下缩短新员工达标周期。
## Atmosphere & style
整场对话务实且节奏明快，兼具高管战略前瞻与呼叫中心一线落地细节。两位嘉宾的互动默契风趣，既有对技术碎片化痛点的犀利剖析，也有对基层员工文化心理转变的温和洞察。听众能够清晰感受到传统客服行业在资本与技术双重驱动下，从被动应对监控焦虑走向主动拥抱人机协同赋能的真实演进脉络。
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) 企业 AI 的核心考量已从单纯“是否采用”转变为部署时机、落地规模以及能否兑现预期的业务成果。
- [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) 语音是最高频且具人性化的交互方式，数据则是 AI 的燃料；低质量数据会被 AI 放大，需结合治理机制与连接能力避免前后端脱节。
- [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) RingCentral 构建覆盖前端 24/7 AI 接线、中端座席无缝上下文转接与事后全量对话智能复盘的端到端客户旅程架构。
- [04:57](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=297s) Avantive Solutions 放弃自建 Avaya 本地大型硬件与机房维护，全面转向 RingCentral 云端架构以敏捷拥抱技术迭代。
- [07:40](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=460s) 在外呼营销受法规严格限制模拟语音的背景下，机器学习专注于预测分析与拨号策略优化，生成式 AI 则聚焦内容生成与行动辅助。
- [07:40](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=460s) 企业摒弃了早期的单一工具碎片化拼凑（Tactical Sprawl），转向以 AI 为核心驱动的顶层企业战略与治理架构。
- [10:45](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=645s) 全量语音分析将 AI 质检覆盖率拉升 174%，并通过客观温和的生成式 AI 反馈表减少人工管理沟通摩擦。
- [10:45](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=645s) 智能体互动培训机器人能够基于质检薄弱环节模拟真实客户场景，形成“质检-反馈-实战演练”的自动化闭环。
- [13:22](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=802s) 座席对 AI 的态度在 36 个月内从抵触转为积极拥抱，温和非评判性的交互方式让员工将 AI 视为赋能工具而非监控手段。
- [16:06](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=966s) 私募股权 3-5 年的投资周期与 AI 技术 12-24 个月的快速代际跃迁产生时间错配，倒逼董事会设立双轨议题并下沉高管 AI KPI 考核。
- [18:10](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=1090s) 当前最具即时投资回报率的落地点是“AI 实时座席辅助”（AI Assist），能显著缩短新员工成熟周期（Speed to green）并提升人效。
- [18:10](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=1090s) 在强监管业务中，AI Assist 通过实时识别确保免责声明逐字合规，并结合事后 AI Conversation Expert 实现毫秒级复盘。
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) 开场与 Agentic Voice AI 理念：从自动化走向全流程赋能
在面对现场 12,000 名与会者的热烈开场中，来自 RingCentral 的全球技术解决方案高级副总裁 Amir Hameed 开启了本场炉边对话。他幽默地调侃了昨天的排队盛况，并直言当下 AI 的发展早已跨越了“是否采用”的阶段，真正的焦点在于“何时部署、部署多少，以及它是否真能带来预期的业务成果”。与其他泛泛讨论 AI 的场景不同，RingCentral 将核心聚焦于**智能体语音 AI（Agentic Voice AI）**。Amir 强调，数据固然是 AI 的燃料，但语音才是最人性化、最直接的交互形式；如果底层数据质量不佳，AI 只会将问题放大，而在语音交互中，对话的连接方式、治理机制与规模化能力决定了一切成败。
现实中很多 AI 项目之所以停滞不前或未能兑现承诺，主要症结在于工具层面的战术性蔓延（tactical sprawl）、前后端数据断层，以及 AI 智能体与人类座席切换时的上下文丢失。Amir 指出，RingCentral 从整个客户旅程的视角构建能力：在前端通过 24/7 在线的 AI 接线员与代表实现高度自动化交互；当需要转接给真人时，借助 Agent Assist（座席辅助）无缝传递所有对话背景，避免让客户反复自报家门；而在通话结束后，通过对话智能（Conversational Intelligence）对每一次交互进行全量分析复盘，从而在不浪费客户宝贵时间的前提下最大化满意度。
```mermaid
flowchart LR
accTitle: Agentic Voice AI Customer Journey Architecture
subgraph FrontEnd["前端自动化接入"]
A["AI 接线员 / AI 代表<br/>(24/7 语音交互 & 意图理解)"]
end
subgraph HumanAssist["座席实时协同"]
B["无缝上下文移交<br/>(数据同步 / 免重复验证)"]
C["Agent Assist 座席辅助<br/>(实时建议 / 客户画像)"]
end
subgraph PostCall["事后智能复盘"]
D["Conversational Intelligence<br/>(全量对话分析 / 体验优化)"]
end
A -->|"需真人介入"| B
B --> C
C --> D
```
- **数据与语音交互**：数据是 AI 的燃料，不良数据会被 AI 放大；而语音承载着最高频且真实的业务与客户沟通。
- **痛点规避**：解决工具碎片化（tactical sprawl）以及前台（front office）与后台（back office）数据割裂的问题。
- **核心理念**：“无论你为谁工作，工牌上写着什么，归根结底我们都是在为客户服务。”
---
### [04:57](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=297s) Avantive Solutions 业务背景与告别传统本地架构的转折点
Avantive Solutions 首席执行官 Frank Pettinato 随后登场，介绍了公司的基本盘。作为一家拥有 30 多年历史的全球业务流程外包（BPO）服务商，Avantive 主要深耕两大垂直领域：医疗与生命科学（Healthcare & Life Sciences）以及电信（Telecommunications）。其中电信业务涉及大量外呼电话营销（outbound telemarketing），这是一个极其依赖数据驱动且充满挑战的环境。尽管法律法规严格禁止在电话外呼营销中直接使用 AI 语音生成，但这并不妨碍他们全面运用 AI 在通话过程中为真人座席提供实时支持。
当被问及当初放弃传统平台的转折点（breaking point）时，Frank 回忆道，大约在 10 年前，公司原先使用的传统 Avaya 本地硬件系统面临升级周期。当时管理层意识到，继续耗费巨资自建和维护“大型本地铁皮机房”（big iron data center）与自研本地部署方案已不再是长久之计。他们果断决定拥抱云端，选择与 RingCentral 合作，将技术基础设施交给具备快速创新能力的云伙伴。Amir 总结道，过去十年行业经历了从本地硬件向云端的大迁移，而如今 AI 则成为了比上云更迅猛的全新催化剂。
- **核心行业聚焦**：医疗与生命科学、电信领域的外呼营销。
- **合规界限**：受外呼电话营销法规限制，前端禁止使用 AI 模拟语音外呼，但支持座席侧的实时 AI 赋能。
- **技术转折**：从自建 Avaya 大型本地硬件全面转向 RingCentral 云原生平台，摆脱硬件折旧与自研维护负担。
---
### [07:40](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=460s) 从战术工具到战略核心：Avantive 的 ML 与 AI 演进之路
谈及 Avantive 在智能化领域的探索历程，Frank 详细拆解了机器学习（ML）与生成式 AI（Generative AI）在其业务中的不同定位。Avantive 早在约五年前就开始对机器学习进行大规模投入，主要将其用于**预测与分析**；而 AI 平台则专用于**内容生成与执行行动**。在外呼营销中，最大的痛点是“接通率”（live contact rates），即如何让不愿意接陌生电话的客户接听。Avantive 将沉淀的存量数据注入 ML 算法，动态指导每天、每小时乃至每一分钟的拨号策略，确保在精准的时间向精准的目标推送最合适的产品，这套系统至今依然稳定高效地运行着。
在生成式 AI 的落地过程中，Frank 分享了一个关键教训：在过去约三年的时间里，公司最初采用的是“战术性打法”，即针对质检问题找一个质检 AI 工具，针对座席辅助找一个辅助工具，针对培训开发再买一个独立工具。虽然这种轻资产模式便于快速更换供应商，但到了 2026 年，这种碎片化的工具拼凑已经无法满足发展需求。Avantive 彻底重塑了顶层战略，将 ML 与 AI 投资上升为核心战略支柱。正如 Frank 所言：“我们实际上是一家恰好在做呼叫中心业务的 AI 公司（We're actually an AI company that happens to be a call center）。”从董事会到高管团队（ELT），核心议题已全面转变为评估 AI 带来的业务风险与利用 AI 提升生产力，同时在严苛的监管红线下严谨对待每一次直接面向消费者的交互。
- **概念区分与应用**：
- **Machine Learning（机器学习）**：用于预测分析与拨号策略优化，解决外呼接通率难题。
- **AI / Generative AI**：用于内容生成、座席实时支撑与行动编排。
- **思维模式跃迁**：从早期“一痛点一工具”的碎片化战术采购，转变为统一的战略级“AI 优先”企业架构。
- **治理与管理重点**：董事会与管理层全面聚焦于 AI 风险控制与生产力释放。
### [10:45](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=645s) AI 质量管理闭环：从 174% 的质检覆盖到智能体模拟实操
在监管动态（如 FCC 对呼叫中心的限制持续收紧）与客户体验诉求不断升级的背景下，企业必须在合规与服务体验之间取得平衡。针对 Amir 提出的“质量监控覆盖率提升 174%”这一关键指标，Frank 深入拆解了 Avantive Solutions 内部的质检落地机制。对于拥有数千名座席的 BPO 团队而言，核心理念是“让每位座席都成为最好的自己”。因此，公司几乎将每一通客户来电都接入语音分析平台，结合 AI 工具深度解析通话与富数据（enriched data），自动化输出质量评分卡（Quality Scorecard）。
传统模式下，AI 生成的评分卡由主管（Manager）人工传达给座席，但实际落地中容易因主管个人的“风格差异”（stylistic issues）导致沟通出现摩擦或压力。为了解决这一痛点，Avantive 部署了生成式 AI（Generative AI）工具，以统一、客观且温和敏锐的语气自动生成 QA 教练反馈表。座席可以直接与该系统进行建设性互动，避免了面对面反馈时的紧张感。随后，系统会根据质检短板自动匹配培训模块，并触发一个智能体交互式培训机器人（Agentic AI interactive training bot）。该机器人能够模拟真实的客户对话场景，针对座席在质检中暴露的具体问题提供高度定制化的针对性实战演练。
```mermaid
flowchart TD
accTitle: AI 质量管理与智能体培训闭环
subgraph QA ["全量分析与智能质检"]
A["全量通话接入语音分析"] --> B["AI 深度解析与富数据提炼"]
B --> C["自动生成质检评分卡"]
end
subgraph Feedback ["温和反馈与赋能"]
C --> D["GenAI 生成客观温和反馈"]
D --> E["自动匹配薄弱项培训模块"]
end
subgraph Training ["智能体实操"]
E --> F["Agentic AI 培训机器人"]
F --> G["模拟定制化客户场景对练"]
end
```
- **全量质检与评分自动化**：将 AI 质检覆盖率大幅拉升 174%，摆脱了传统人工抽检覆盖不足的瓶颈。
- **消除人际沟通摩擦**：用规范温和的生成式 AI 反馈替代部分直接的人工主管施压，创造低心理负担的改进环境。
- **智能体实战对练**：培训机器人根据质检反馈动态生成专属客户交互场景，实现“质检-辅导-实操演练”的自动化闭环。
---
### [13:22](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=802s) 破除“被监控”焦虑：36 个月内座席对 AI 反馈态度的文化转变
作为依赖人力资产的 BPO 企业，如何打消基层员工对“被 AI 严密监控”的抵触情绪并促成技术采纳，是一个至关重要的课题。Frank 回顾了过去 36 个月的剧烈变化：在 3 年前首次部署座席端 AI 工具时，由于当时 AI 概念尚未普及，内部遭遇了极为明显的抵触与推力；然而快进到今天，座席群体的态度发生了彻底逆转，许多人甚至更倾向于先与机器系统进行交互，而非直接面对人类主管。
当 Amir 追问这一转变是源于高流动性带来的年轻一代数字化原住民，还是由于现有员工被逐步培养起来时，Frank 认为两者兼有，但更关键的是组织文化与工具价值的演进。呼叫中心本身流动性较高，但 AI 工具被刻意设计成以“温暖、友好且具建设性”的方式提供建议。在整体流程中依然保留了“人机协同”（Human in the loop），团队主管仍会进行一对一的辅导补充。更重要的是，现代人已经习惯在日常生活或健康管理中向 AI 寻求建议来改善自我，将这种体验延伸到呼叫中心后，员工意识到 AI 不是惩罚工具，而是帮助自己达成最佳工作表现的赋能平台。
- **36 个月的心态演进**：从 3 年前对 AI 的明显抵触与质疑，转变为如今积极拥抱并偏好机器提供的初审反馈。
- **温和非评判性交互**：AI 消除了解读偏差与情绪压力，以中立、支持性的口吻提供改进建议。
- **人机协同（Human in the loop）**：AI 负责高频、客观的数据复盘与日常练习，人类教练负责高阶一对一沟通与情感支持。
---
### [16:06](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=966s) 私募股权驱动下的 AI 战略：技术迭代与 3-5 年投资周期的博弈
在企业战略层面，Avantive Solutions 已经从战术性解决单一问题，转变为将自身定位为一家“以 AI 为核心驱动”的公司。作为一家由私募股权（Private Equity, PE）控股的企业，董事会对 AI 的关注已上升为强制性议题：每次董事会都会专门开辟核心议程，前半部分深入研判市场中新涌现的 AI 技术可能带来的商业颠覆与潜在风险，后半部分则评估 AI 在生产力提升、产出增加与降本增效方面的具体收益。
这种战略推进面临着严峻的周期错配挑战——私募股权基金的典型投资与退出周期为 3 到 5 年，然而 AI 技术在过去 12 到 24 个月内的演进步伐已经远远超越了传统资本的评估窗口，给投资人与管理层带来了巨大的紧迫感。为了全面落地，Avantive 执行领导团队（ELT）的每位高管都背负着与 AI 部署与激活直接挂钩的专属 KPI，涵盖财务会计系统、人力资源信息系统（HRIS）、招聘流程等各个后台与日常运营环节。公司自上而下明确传达一个核心原则：AI 的目的绝非取代员工的饭碗，而是作为赋能杠杆，消除重复枯燥的工作，让每位员工成为更高效的自己。
- **董事会双轨议题**：定期设立专项讨论，一边严格审查新兴 AI 技术对业务构成的颠覆性风险，一边追踪生产力与成本优化的实际收益。
- **投资周期与技术速度脱节**：PE 机构 3-5 年的投资规划与 AI 技术每 12-24 个月一次的代际跃迁产生时间错配，倒逼战略加速落地。
- **全员管理层 KPI 考核**：将 AI 激活指标强制分解至所有业务线（HRIS、会计、招聘等），推动非核心与繁琐流程的全面智能化。
---
### [18:10](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=1090s) 智能体前沿落地：实时座席辅助（AI Assist）与高合规场景演进
谈及当前 Agentic AI 的前沿探索，Frank 明确指出，尽管业内对完全自主的“独立 AI 智能体”（AI Agents）讨论热烈且其确有应用场景，但目前最具商业价值和立竿见影回报的切入点是**“AI 实时辅助座席”（AI Assist）**。这一方案既能显著缩短新员工从入职到达到合格绩效标准的周期（即“Speed to green”），又能持续拉动在职成熟座席的单位产能，具有无可辩驳的投资回报率。
在金融、医疗等高度受监管的严苛合规业务中，法律明确要求座席必须在特定时机逐字逐句（verbatim）念出合规免责声明和固定话术，次序不得出现差错。AI Assist 能够在通话过程中提供实时纠偏与话术提示；通话刚一挂断，借助 RingCentral 的 AI Conversation Expert（原 RingSense 平台），座席即可在毫秒级内获得详尽的实时通话复盘与行动建议，随后再无缝流转进入深度质量管理流水线。Amir 总结强调，这些融合了实时辅助、事后分析与 Agentic 架构的能力并非停留于 PPT 上的概念，而是已经在生产环境中稳定运行的成熟方案。
- **缩短成熟周期（Speed to green）**：AI Assist 成为新员工入职后的实时“副驾驶”，大幅降低带教成本并加速产能释放。
- **严苛合规精准把控**：在法律严格要求逐字合规的场景下，通过实时语义识别确保合规声明与关键流程 100% 准确执行。
- **实时与事后双重闭环**：通话中提供实时线索辅助，通话结束后由 AI Conversation Expert 立即交付结构化复盘与改进计划。
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) 何时部署、部署多少，以及它是否真能带来预期的业务成果 — Amir Hameed
> [00:04](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=4s) 无论你为谁工作，工牌上写着什么，归根结底我们都是在为客户服务。 — Amir Hameed
> [04:57](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=297s) 大型本地铁皮机房 — Frank Pettinato
> [07:40](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=460s) 我们实际上是一家恰好在做呼叫中心业务的 AI 公司（We're actually an AI company that happens to be a call center） — Frank Pettinato
> [10:45](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=645s) 让每位座席都成为最好的自己 — Frank Pettinato
> [13:22](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=802s) 温暖、友好且具建设性 — Frank Pettinato
> [16:06](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=966s) AI 的目的绝非取代员工的饭碗，而是作为赋能杠杆，消除重复枯燥的工作，让每位员工成为更高效的自己。 — Frank Pettinato
> [18:10](https://www.youtube.com/watch?v=QBi1Cz1_DJU&t=1090s) AI 实时辅助座席（AI Assist） — Frank Pettinato
## Mentioned
Amir Hameed, Frank Pettinato, RingCentral, Avantive Solutions, Ai4, Agentic Voice AI, Agent Assist, Conversational Intelligence, Avaya, FCC, AI Conversation Expert, RingSense
## Video description
Amir Hameed - Senior Vice President - Global Technical Solutions, RingCentral
Frank Pettinato - CEO, Avantive Solutions
Most enterprises have automated customer interactions, yet few have orchestrated them. Agentic Voice AI goes beyond automation, reasoning across conversations, enterprise data, workflows, and human expertise to achieve outcomes. This session explores how enterprises move from AI agents that respond to requests to agentic systems that orchestrate intelligent customer engagement across voice and digital channels. Drawing on real-world deployments, we'll examine the architectural decisions, governance, and operational practices that distinguish successful production deployments from stalled pilots. Attendees will leave with actionable frameworks and scaling considerations you can apply immediately.
[Source](https://www.youtube.com/watch?v=QBi1Cz1_DJU)
Source quotes
[0:04] Amir Hameed: 何时部署、部署多少，以及它是否真能带来预期的业务成果
[0:04] Amir Hameed: 无论你为谁工作，工牌上写着什么，归根结底我们都是在为客户服务。
[4:57] Frank Pettinato: 大型本地铁皮机房
[7:40] Frank Pettinato: 我们实际上是一家恰好在做呼叫中心业务的 AI 公司（We're actually an AI company that happens to be a call center）
[10:45] Frank Pettinato: 让每位座席都成为最好的自己
[13:22] Frank Pettinato: 温暖、友好且具建设性
[16:06] Frank Pettinato: AI 的目的绝非取代员工的饭碗，而是作为赋能杠杆，消除重复枯燥的工作，让每位员工成为更高效的自己。
[18:10] Frank Pettinato: AI 实时辅助座席（AI Assist）
Source quotes
[0:04] Amir Hameed: when, how much, and whether it actually produces business outcomes.
[0:04] Amir Hameed: data is the fuel for AI
[4:57] Frank Pettinato: big iron data center business
[7:49] Frank Pettinato: an AI company that happens to be a call center.
[13:22] Frank Pettinato: become the best version of themselves
[18:10] Frank Pettinato: speed to green
Original transcript
[0:04] Excellent. Thank you. How's everybody doing?
Been a good event? 12,000 people, not bad.
[0:11] I hope you guys didn't have to wait in line
yesterday. Some of you did, unfortunate,
[0:15] but we didn't know what to expect. Listen,
it's a pleasure to be here with you all today.
[0:19] Um, I just wanted to kind of start off and set
the tone. So, I'm with RingCentral. I lead global
[0:25] uh technical solutions, which means I get to
talk to customers every day around the world
[0:30] and really understand some of the challenges that
they're going through and how we can apply AI to
[0:37] uh to help them alleviate that and truly drive
business outcomes. So, I have the pleasure of
[0:40] having Frank with me. He's a CEO of Avantive
Solutions and one of our dear customers and really
[0:47] the the lion's share of the talk is going to be a
little fireside chat with him. But, before we do,
[0:52] I just wanted to kind of set the tone. You know,
clearly we're all here, 12,000 people jammed in.
[0:58] AI is a reality. There's no question about
it, right? AI is here. It's not a matter of
[1:04] if, but when and how much AI are you
going to deploy? And the reality is,
[1:09] you know, the AI that you deploy, is it really
going to drive the outcomes that you're expecting,
[1:14] right? So, I think we're we're beyond
that. And, you know, at RingCentral,
[1:18] we're we're kind of focused on a little different
angle slightly. We're focused on agentic voice AI.
[1:25] Right? So, a little bit different perhaps
from some of the other presentations that
[1:28] you're talking about because we know,
you know, uh data is the fuel for AI.
[1:33] Right? Data is the fuel. So, if you get bad data,
it's it's going to just amplify with the AI.
[1:37] But, one of the things we're focused in on is the
actual voice, right? Cuz that's the most human
[1:42] form of AI. So, whether you're conversing with,
you know, a business partner, with colleagues at
[1:49] work itself, or more importantly, arguably me, or
arguably with your customers, right? So, that's
[1:54] the type of angle that we're talking about. and
we're looking at the conversations themselves and
[2:00] how people are connecting, how they're governing,
and how they're scaling. So, we're going to talk
[2:04] to you a little bit about that today, but we also
realize that not all AI projects are a success.
[2:11] There's a lot of stall. There's a lot of uh hey,
wait a second. I was kind of sold this idea,
[2:16] but I didn't really get the benefits or
the business outcomes that I was promised.
[2:21] And there's various reasons for all of that
that you can see here. You know, tactical
[2:27] sprawl from the tools perspective, you know, if
you don't have all of them in the right spot,
[2:31] you know, that What about the context? You know,
you're you're you're mixing now. You've got AI
[2:36] agents. You've got human agents. Do you have
a seamless handoff? Or does Do people need to
[2:41] repeat themselves, re-identify themselves? You
know, these are something that you would expect,
[2:46] but you'd be surprised it's not always there.
And again, data layers, right? Front office,
[2:51] back office. Are they connected? Are they sharing
all the valuable information that they have,
[2:57] or is it being lost somewhere in the conversation?
And so again, at at RingCentral, we've been we
[3:02] kind of look at it as as a customer journey,
right? We're looking at it from a customer lens.
[3:06] And so up front, we try to automate as much as
possible. We leverage uh you know, things such as
[3:12] AI receptionists, AI representatives.
You know, these are the tools where
[3:17] you're leveraging voice to access the business,
your provider, whomever they may be, but
[3:24] you can automate it. People are you know,
the agents can be available 24 by 7, very
[3:28] easy to set up. So, we try to automate as much as
possible, give the customers the information that
[3:34] they need leveraging AI very quickly. Now,
when they do need to get to an individual,
[3:40] we also want to leverage AI, right? So so so
we have things such as our agent assist where,
[3:47] you know, we know so much about the customer
already, we want to make sure that we empower
[3:52] the agent with that information. Now, some
of that information may have been shared
[3:56] with the digital agent that has to seamlessly
transfer over to that human agent so that again
[4:03] nobody's time is being wasted, right? You you've
got the information, the context that you need
[4:08] to have a meaningful conversation with that
customer. And especially if they're a you know,
[4:13] a platinum level high roller like Frank here, you
know, you want to make sure that we're giving him
[4:17] what he needs and we're not wasting his time. And
then of course analyzing after the fact, right?
[4:23] This is where the conversational intelligence
comes in. We're leveraging AI after the fact to
[4:28] glean not just some but every single conversation,
every single interaction at scale to ensure that
[4:37] we hit the mark and delight customers cuz at the
end of the day it doesn't matter who you work for,
[4:42] doesn't matter what your badge says, we all
work for the customer one way or another. And
[4:47] with that Frank, I wanted to kind of share with
you. You've got some some tremendous experiences.
[4:53] You know, why don't you start off by introducing
yourself and your role and in your company?
[4:57] Yeah, thanks Amir. Again, Frank Petanado the
CEO of Avantive Solutions. We're global BPO
[5:03] primarily focused on two verticals, healthcare
and life sciences and telecommunications.
[5:09] In particular our telecommunications sector is an
area that we deploy a lot of machine learning and
[5:14] AI technology as a lot of our work is outbound
telemarketing. That is a very data intensive
[5:21] um difficult and challenging environment. So we
definitely see an opportunity to to leverage that.
[5:28] Now as many people in the room may
know, there's regulatory prohibitions
[5:32] on the use of AI voice in outbound telemarketing,
but that doesn't prohibit us from using AI
[5:38] much in the examples you gave you to support the
agents during the during that call interaction.
[5:44] So again, thank you for having me today.
Yeah, no, it's a pleasure. So listen, I think
[5:48] and and this is why we appreciate when we get
some time to to talk to an audience. I love
[5:52] to have a customer because many of you may be
in the same shoes that that Frank has been in.
[5:57] And one of the things that's that's interesting
to me is, you know, can you talk to us about your
[6:01] breaking point? Like, you know, obviously you've
been there for some time. The company's been there
[6:06] for some time. You've had legacy platforms that
you were on. But, what caused you to say, "Hey,
[6:11] now it's time to move off that legacy."
What was that breaking point? Yeah,
[6:15] great question. So, company's been around for
30-plus years. Uh we've been with RingCentral
[6:21] roughly 10 years now. Uh previous to that, we've
been on we were on a legacy of Avaya platform.
[6:28] That platform required an upgrade. Um you
know, we just woke up and decided we're
[6:34] not going to be in the big iron data center
business. That wasn't what we're going to do.
[6:38] We needed a partner that could give us the
technology and services we need, but also
[6:43] bring us along on a pretty aggressive innovation
cycle. Of self-funding or developing premise-based
[6:49] solutions was not uh a long-term strategy. And,
you know, many other call centers, you know,
[6:55] basically woke up around that same time and and
moved to the cloud. Um our decision was really
[7:01] forced upon us foisted upon us, you could argue,
simply because we had a we met an inflection point
[7:08] of you know, massive investments. Massive
investments and you recognized the legacy platform
[7:12] wasn't going to do it for you anymore. And and
that's very common. Listen, in our business,
[7:16] you know, the the first shift I like to think was
coming to the cloud, right? You've got the legacy
[7:20] iron as you'd called it. You realize you've
you've sweat that asset. It's time to move on.
[7:25] The first step was moving to the cloud
and we saw a huge shift to the cloud,
[7:29] but I would argue now with AI, it's a catalyst.
It's driving so much more than than what we
[7:36] that what we saw before. So, now, look, you guys
have been kind of like early adopters of AI,
[7:40] right? So, can you talk to me a little bit
about your journey? How you guys dabbled in it?
[7:44] What were some of your earliest learnings
and and where you're taking it now? Yeah,
[7:49] we'd say if I went back to about 5 years ago, we
started fairly significant investments in machine
[7:56] learning. And for context, the way I'm going to
discuss the difference between machine learning
[8:01] and AI, we use our machine learning platforms
for predict predictions and analytics.
[8:07] We use our AI platforms to generate content
or generative energetic actions taken. Two
[8:14] separate kind of thought processes, two separate
solutions, two separate inputs and outputs, but
[8:20] you for this discussion, I wanted to make
sure that there was a definition that I used.
[8:24] Uh [clears throat] in outbound telemarketing,
some of you may know or not know,
[8:30] we're battling live contact rates. People not
wanting to answer their phone, but our job is
[8:35] to get people to answer the phone and then sell
services that we feel will be valuable to them.
[8:41] So, the the basic premise was if we
could take the data we already have,
[8:46] put it into a machine learning algorithm of sorts,
the output would help us be more strategic on
[8:52] the approach that we would take day-to-day,
hour by hour, in some cases minute by minute
[8:58] on our dialing strategy to ensure that we're
contacting the right people at the right time
[9:02] with the right offer. Um early adopter, I'd say is
probably 5 years ago is probably fair statement.
[9:10] Those platforms have had very little change over
that period of time and they've worked famously.
[9:15] We're going through another kind of upgrade at
this point to enhance those services. As far as
[9:21] AI in particular, um this is maybe one of the big
learnings that I want to share with the group is
[9:28] we started off using AI tools as tactical
solutions to problems. So, for example,
[9:36] we needed to solve a quality management issue, we
would look for a AI tool. We needed to solve an
[9:42] agent assist problem, we look for a specific tool.
We want to solve a training and development issue,
[9:49] we look for a specific tool. So, that worked
pretty well over the last 3 years or so.
[9:55] We were fortunate because we didn't make long-term
investments, we could switch from vendor to
[10:00] vendor. But now here in 2026, we look at We're
looking at it very differently. We're looking at
[10:07] both our machine learning and our AI investments
as a strategic initiatives. And the core strategy
[10:13] of the business is fundamentally changed. And
the way I think about it is we're actually an AI
[10:18] company that happens to be a call center. That's
a good point. Uh from our board to our ELT, the
[10:23] topics are risks from introduced to the business
from AI and opportunities to drive productivity
[10:31] from AI. But you had mentioned it earlier,
at the end of the day, the person who pays
[10:36] the bill is the consumer. 100%. So, the
interactions that we put forth and the
[10:40] way we support the consumer, we're very sensitive
to, especially given the regulatory environment,
[10:45] and we feel the FCC's actually swinging further in
restrictions. So, there's a balancing act and the
[10:51] dynamics are playing out in real time on AI in the
call center, we'll say, or customer experience.
[10:57] But at the end of the day, we're all here to serve
the serve the consumer, and the tools that we are
[11:02] now deploying and the strategy that we have, we
believe is really going to ultimately help the
[11:08] consumer get a better service from their from
their customers. Perfect. So, hey, let's let's
[11:13] peel the onion on one thing. So, you mentioned
AI quality quality management, right? You know,
[11:19] in some of our discussions, we talked I think you
were saying about a 174% more quality monitoring.
[11:26] What does that actually mean for your agents, your
your supervisors, and ultimately for your clients?
[11:32] So, in the call center, we have, you know,
thousands of call center agents. Each one of them
[11:38] deserves to be the best version of themselves. So,
because of that, we really focus on coaching and
[11:44] development. We take virtually every single call,
we put it through our voice analytics platform.
[11:52] That voice analytics platform then goes into our
quality management program. We use AI tools to
[11:58] analyze the call, the enriched data, and
then come out with a quality scorecard.
[12:03] So, up until recently, that quality scorecard was
then given to the manager Yeah. to deliver that
[12:08] feedback to the agent. But, because we were able
to do this using AI, we're able to generate more
[12:13] quality scorecards. So, we're able to provide more
meaningful feedback to the agents and more mean-
[12:19] meaningful feedback to the uh team managers.
So, we thought that was working really well,
[12:25] but then we started to get a sense that, "Hey,
maybe the delivery of a human to a human Yes.
[12:31] was causing some cuz people have stylistic
issues." Yeah. Yeah. So, we deployed a uh uh um
[12:38] a generative AI tool to basically come
up with a QA coaching feedback Yeah.
[12:45] in a very consistent but sensitive form. So,
there you're giving feedback to the agent.
[12:51] They can interact with the form in a constructive
manner, so they don't feel that pressure of
[12:59] one-to-one feedback. Right. Right. Right. And then
at the end, it's then assigning training modules.
[13:05] Those training modules are then developed and
delivered through an agentic AI interactive
[13:11] training bot that replicates an interaction with
either uh with a customer. That's tailored to the
[13:18] exact feedback that they got during the during the
coaching session. All right. So, so look, you guys
[13:22] are a BPO, so obviously people are your strength,
right? You want to keep your people, you want to
[13:26] train your people, you want to retain your people
cuz ultimately that that's your touch point to get
[13:32] more business and keep customers happy. Now, in
a lot of cases people are concerned, right? Like,
[13:37] oh, you're going to bring AI. AI is monitoring me.
So, how do you kind of overcome that um concern
[13:44] from the staff to get them to truly embrace the
value of the AI tools that you're bringing forth?
[13:50] I've seen a shift in 36 months. So, our initial
agent-based AI tools were deployed 3 years ago.
[13:59] And we saw very meaningful and clear disruption
and pushback from the agents. Because
[14:09] argue on the time frame AI wasn't a thing then,
we'll call it, okay? So, now if we fast forward to
[14:15] today, the generation of a call center agents we
have are very comfortable with it. Prefer the um
[14:25] interaction with a machine versus a person.
just to just to just interrupt you there,
[14:29] is that because the staff has turned over and
perhaps you've got a younger, more AI-savvy pool
[14:35] of talent or is it the folks that that kind of
trained and and came along with the for the ride?
[14:40] I I think it's cultural, like, you know, the the
adoption or the be able for for people like a call
[14:46] center agent to understand the value of the tool
has has evolved over 36 months rather quickly. We
[14:52] we've seen it in our own call centers. Of course
there's some turnover. Call centers have a lot of
[14:56] turnover. Sure. Um but right now I would say
um if you asked our call center agents, they
[15:02] actually prefer the interactions with the AI. But
the feedback is is given in a warm and friendly
[15:09] constructive way, right? We've designed it that
way. But um it is still augmented by one-to-one
[15:16] feedback from a from a coach or a leader. So,
there's still a human in the loop. There is.
[15:21] Yeah. Yeah. But the acceptance of the technology,
the acceptance that you're getting feedback from
[15:26] a machine, um in my opinion, has has grown. But
equally important, you had mentioned it earlier,
[15:36] everybody wants to be the best version of
themselves. Sure. I'm sure everybody in this room
[15:39] is asked some version of an AI about a question
about their personal life or their health or
[15:45] whatever and gotten some feedback to try
to become the best version of yourselves.
[15:50] And we basically are providing that
platform to our call center agents.
[15:54] And the outcome is they become the best agent
they can be. Benefits us as their employer and
[16:00] the service we provide for our customers and our
customers' consumers. Makes sense. So, hey, look,
[16:06] you talked about moving from solving tactical
problems to building a company strategy around
[16:10] AI. In fact, you even said that you feel you
guys are an AI company first and everything
[16:14] else comes along with it. Um so, what does
it actually look like at Alvantive right now?
[16:19] So, based on this year, uh board directives, we're
private equity owned company. Board directives are
[16:27] a certain portion of our board meetings
are directly related to AI risks,
[16:32] specifically what's emerging in the marketplace
that could be a risk to the business, and then AI
[16:39] benefits, we'll call it. You know, what
productivity increases, what production increases,
[16:45] what cost savings, what can we put in place? And
they're oddly enough, uh weighty sub- a weighty
[16:52] subject and typically early on. Um as you guys can
imagine, anybody investing money in a business,
[16:58] any business, doesn't really matter,
something as boring as a call center,
[17:02] is looking at what the risk of that investment's
going to be in 1 year or 2 year based on emerging
[17:08] technology. Think about how quickly we've
come over just 24 months, let alone 12 months.
[17:14] You know, and uh private equity firms are
making three to five-year investments.
[17:19] So, you could see the window of technology is
moving faster than the investment windows. And
[17:24] that's causing a little bit of stress. No doubt.
There's so much unknown and again the torrid pace
[17:30] of innovation is is incredible. And then at
the ELT level, each one of my leaders is given
[17:36] KPIs that specifically revolve around deployment
and activation of AI tools and technology
[17:44] everything from our accounting systems to our
HRIS systems, recruiting you know, any any mundane
[17:50] function that's happening within the organization
is ripe for an opportunity. And you know,
[17:56] obviously there's a cultural issue. We have to be
sensitive that we're not replacing people's jobs.
[18:01] We're basically telling everyone this is a
tool to make you a better version of yourself
[18:06] and to drive performance and productivity for
the company. Absolutely. We got a little bit of
[18:10] time left. So, look, you guys are on an agentic
frontier, no question. Um you know, what are the
[18:16] capabilities you're most focused in on right now?
Like if you kind of looking forward based on what
[18:21] you know today and what you see? Wha- where's
your where's your head at? We believe although
[18:29] there's a lot of talk about AI agents and I think
there's a place for it in the marketplace, AI
[18:34] assist of existing agents I think is really going
to be a huge value proposition. Uh we could take
[18:42] it's speed of green for new agents, it's increased
performance and productivity of existing agents,
[18:48] the cost is is justified. And then us, we work
in a highly compliant environment. So, verbatim
[18:54] speaking specific words and specific sequences
are critical um and legally required. Absolutely.
[19:01] Yeah. Getting that getting that inter- interaction
in real time and then post-call real time
[19:09] um using RingSense that we have we
rebranded it. What's it called? Yeah, yeah,
[19:14] AI conversational expert. Yeah, sorry, excuse my
AI conversational expert. Yeah. So, the agent's
[19:19] getting immediate feedback right after the call.
Right after the call. Yeah. Then it goes through
[19:23] that quality management process that we talked
about where they're getting more rich and detailed
[19:27] feedback with an action plan on how to improve.
Got it. So, hey, listen, last few minutes here,
[19:33] we we talked last few seconds here. We talked
a lot about various solutions, where Frank's
[19:37] head's at. I would encourage you all if you have
the time, come out to booth 1025, RingCentral.
[19:42] You can actually see all of these AI agents, you
can create AI agents, you can put them to use,
[19:47] you can speak with them. You can talk about the
agent assist. Basically, all of the things that
[19:52] we kind of touched on are not features, they're
they're available today. And if anyone is in that
[19:58] path where they're thinking about where
they're going to go to next, specifically
[20:02] as it relates to Aigentic voice AI, I encourage
you to come out to the to the RingCentral booth.
[20:08] That's booth 1025. Thank you so much for
listening. All right. Nice job, Vern. Thank you.
