TaskUs Backs Voice AI as Real-Time Calls Yield Triple Chatbot ROI
Ai4
Voice AI delivers three times the return on investment of commoditized text bots because live spoken interactions accelerate top-line revenue conversion rather than just cutting labor costs.
## Background
Published by Ai4 on YouTube on September 13, 2025, this video documents a fireside chat featuring Sahil Mehta of Regal AI and Jarrod Johnson of TaskUs. According to the description, the session explores how enterprise-grade voice AI agents are moving from pilot experiments into production across high-volume BPO environments, driving lead conversion and operational efficiency.
## TL;DR
TaskUs and Regal AI discuss the operational shift toward deploying autonomous voice agents across enterprise contact centers, contrasting low-latency voice architecture with commoditized text chatbots. The speakers emphasize that voice interactions yield roughly three times higher conversion ROI than messaging while unlocking complete, 100% interaction telemetry for continuous quality assurance. Using a healthcare marketplace case study, they illustrate how proactive outbound voice agents eliminate scheduling drop-offs to drive top-line revenue rather than merely trimming labor costs. Finally, they stress that successful long-term scaling requires rigorous cross-functional alignment between operations and tech, constant knowledge base hygiene, and robust enterprise governance over SMB point solutions.
## Atmosphere & style
The discussion carries a relaxed, collegial, and candid conference tone, opening with lighthearted banter about braving the Las Vegas desert heat. Jarrod Johnson offers the pragmatic, battle-tested perspective of a global BPO leader unafraid to disrupt legacy headcount models, complemented by Sahil Mehta’s grounded focus on telephony latency and unit economics. The exchange feels less like an idealized software pitch and more like an authentic, behind-the-scenes debrief between industry veterans sharing the real operational friction, hygiene challenges, and strategic lessons of scaling enterprise AI.
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=rEGgNLAMocM&t=4s) TaskUs leveraged its experience training foundation models for major tech companies to pivot from defending legacy labor BPO models to launching an agentic AI consulting practice.
- [00:04](https://www.youtube.com/watch?v=rEGgNLAMocM&t=4s) Regal's enterprise platform architecture relies on three core pillars: deep enterprise integrations, context-embedded real-time triggers, and pre-deployment synthetic call simulations.
- [04:08](https://www.youtube.com/watch?v=rEGgNLAMocM&t=248s) Unlike traditional contact centers that manually sample 5% to 10% of calls for QA, voice AI natively analyzes 100% of interaction telemetry to eliminate human variance and test policy changes.
- [08:09](https://www.youtube.com/watch?v=rEGgNLAMocM&t=489s) Service providers face enterprise resistance when trying to sell proprietary core models, making structured partnerships with dedicated software infrastructure providers like Regal essential.
- [08:09](https://www.youtube.com/watch?v=rEGgNLAMocM&t=489s) While text chatbots tolerate multi-second latency and face commoditization, low-latency conversational voice AI drives higher empathy and yields roughly three times the ROI.
- [13:28](https://www.youtube.com/watch?v=rEGgNLAMocM&t=808s) Deploying proactive outbound voice agents in healthcare marketplaces resolves scheduling bottlenecks and cancellations, shifting the value proposition from labor reduction to top-line revenue acceleration.
- [14:52](https://www.youtube.com/watch?v=rEGgNLAMocM&t=892s) The primary cause of failure in enterprise voice AI projects is a disconnect between operational frontline leaders and product or engineering teams.
- [14:52](https://www.youtube.com/watch?v=rEGgNLAMocM&t=892s) Organizations consistently underestimate the ongoing maintenance and knowledge base hygiene required to keep voice agents aligned with constantly evolving enterprise standard operating procedures.
- [16:31](https://www.youtube.com/watch?v=rEGgNLAMocM&t=991s) Enterprise-grade voice AI fundamentally differs from niche SMB point solutions by handling massive scale, complex backend ontologies, and strict governance across regulated industries like fintech, insurance, and healthcare.
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=rEGgNLAMocM&t=4s) Introductions and Organizational Backgrounds
At the Ai4 conference in Las Vegas, Sahil Mehta, Chief Operating Officer at Regal, opened the fireside chat alongside Jarrod Johnson, Chief Customer Officer at TaskUs, welcoming attendees who braved the desert heat to explore complex, real-world enterprise deployments of agentic AI. Jarrod—joking that Las Vegas was "the only place on earth that's hotter than Dallas, Texas right now"—set the stage by outlining TaskUs's footprint. With roughly 15 years in business and a 60,000-person global workforce, TaskUs delivers business process outsourcing (BPO) services across three core pillars: classic customer experience (CX) and contact center operations (60%), trust, safety, risk, and fraud compliance (20%), and specialized AI services (20%). In that AI services wing, TaskUs spent years training and refining machine learning and foundation models for major Silicon Valley tech giants, giving them an early ringside seat to the generative AI revolution before launching their dedicated Agentic AI consulting practice 18 months prior.
Sahil shared the origins of Regal, noting that the company’s founders previously led a 30,000-person contact center organization at home services platform Angi. They observed firsthand that while voice remained the most powerful, ROI-positive customer interaction channel, it had historically been cost-prohibitive and operationally difficult to scale. Regal was built to solve that scaling bottleneck. Backed by over $80 million in funding, the platform has powered more than 250 million calls across hundreds of enterprise deployments. Sahil underscored three primary architectural and operational pillars that set Regal apart from typical market offerings:
- **Enterprise-Grade Infrastructure & Services:** Partnering deeply with systems integrators and service providers like TaskUs to integrate directly with legacy customer data systems, complex ontologies, and enterprise backend workflows.
- **Context-Embedded Event Triggers:** A rules engine capable of dynamically triggering personalized outbound or inbound voice agents based on real-time customer and lead behaviors rather than generic dialers.
- **Enterprise Observability & Pre-Deployment Simulation:** Moving beyond standard conversational intelligence, call transcription, and QA scorecards by running thousands of synthetic conversation simulations to stress-test policies before going live.
```mermaid
flowchart TD
accTitle: Regal Enterprise Voice AI Platform Architecture
subgraph Triggers["Event & Context Layer"]
A["Customer Behavioral Signals / CRM Data"] --> B["Rules-Based Event Engine"]
end
subgraph CoreEngine["Regal AI Platform"]
B --> C["Context-Embedded Voice Agent"]
D["Enterprise Ontology & Workflows"] --> C
E["Enterprise Knowledge Base / RAG"] --> C
end
subgraph Observability["Quality & Observability"]
C --> F["Pre-Call Simulation Engine\n(Thousands of Synthetic Runs)"]
C --> G["Real-Time Call Execution"]
G --> H["100% QA Transcription & Scorecards"]
end
```
[04:08](https://www.youtube.com/watch?v=rEGgNLAMocM&t=248s) The BPO Pivot and Driving Quality at Scale with Voice AI
Addressing why a traditional labor-driven BPO would aggressively champion automated voice agents, Jarrod explained that TaskUs’s client roster is heavily concentrated in high-growth Silicon Valley tech companies and mobile apps. Being "guilty by association" with the world's most tech-forward brands forced TaskUs to anticipate macro shifts. While helping frontier lab clients build and tune foundation models, leadership recognized that these capabilities would rapidly transform consumer-facing operations. Rather than defending legacy headcount models, TaskUs pivoted its internal technical talent to build an advisory practice helping clients implement agentic architectures directly into their support operations.
The conversation turned to customer experience quality. Historically, elite CX required expensive human capital strategies—extensive onboarding, deep training, and long-term retention. Jarrod argued that voice AI introduces a superior quality paradigm because of complete data access. In traditional contact centers, QA managers manually review a mere 5% to 10% sample of calls to gauge sentiment, policy adherence, and product feedback. In contrast, agentic AI platforms natively ingest, process, and analyze 100% of interaction data continuously. This enables organizations to run controlled A/B tests, simulate policy adjustments, and achieve consistent execution at scale, avoiding the inevitable fatigue and variance inherent in large human teams.
- **Sampled vs. Universal QA:** Legacy operations evaluate 5–10% of calls; agentic deployments natively sit on top of 100% of interaction telemetry.
- **Flawless Consistency at Scale:** Voice AI eliminates human variance across edge-case policy enforcement, provided the underlying workflow and knowledge graphs are properly defined.
- **Rapid Policy Iteration:** Enterprises can simulate thousands of downstream customer conversations prior to rolling out process or pricing changes in production.
[08:09](https://www.youtube.com/watch?v=rEGgNLAMocM&t=489s) Selecting Best-of-Breed Partners: Why Regal and the Primacy of Voice
Jarrod recounted TaskUs's internal evolution toward vendor selection. Around two years prior, TaskUs attempted to develop proprietary AI technology in-house, but quickly realized the limits of their identity: "We decided we would go create the AI technology ourselves, and then we forgot we were a services company." While TaskUs excelled at creating workflow wrappers and agent-assist tools, enterprise buyers refused to purchase core generative models from non-software companies lacking dedicated engineering infrastructure. Consequently, TaskUs initiated a structured market evaluation, screening roughly 800 funded AI startups in the customer service space down to two dozen candidates. Their RFI prioritized robust technological fundamentals, referenceable enterprise case studies with proven ROI, long-term product vision, and cultural partnership alignment, while treating initial pricing as secondary.
| Evaluation Dimension | Legacy / Competitor Landscape | TaskUs Selection Criteria (Regal AI) |
|:--- |:--- |:--- |
| **Primary Modality** | 90%+ text chatbots, messaging widgets, and IVR deflection | Native, low-latency, conversational Voice AI |
| **Domain Expertise** | Web-chat wrappers bolted onto CRM suites | Deep CCaaS, telephony, and contact center stack heritage |
| **Track Record** | Unproven roadmaps ("voice is coming next quarter") | Active voice partnership dating back to 2019 |
| **Enterprise Readiness** | Generic API wrappers without governance | Simulation testing, deep ontology mapping, and telephony integration |
Jarrod highlighted that while nearly every competitor was pitching text-based chatbots or treating voice as a vague roadmap promise, Regal was already purpose-built for low-latency voice. Sahil elaborated that the industry had long rationalized chatbots as superior when they were merely cheaper, tolerating 15-second response delays that would completely derail a live voice conversation. While text-based agents are increasingly commoditized by platforms like Salesforce Agentforce, voice interactions deliver significantly higher customer empathy and convert at roughly three times the ROI despite a higher per-interaction compute cost.
[13:28](https://www.youtube.com/watch?v=rEGgNLAMocM&t=808s) Real-World Enterprise Deployments: Healthcare Lead Conversion
Moving from theory to production case studies, Jarrod described a deployment for a venture-backed mental wellness technology platform that operates a two-sided marketplace connecting patients with licensed therapists and counselors. While digital marketing campaigns drove substantial inbound patient interest, the company suffered catastrophic drop-off before patients attended their initial consultation. Scheduling friction, last-minute cancellations, provider unavailability, and mismatched calendar slots created a severe conversion bottleneck that severely restricted top-line revenue.
TaskUs and Regal implemented proactive outbound voice agents to engage registered users immediately upon drop-off, navigate complex scheduling calendars, reschedule canceled appointments, and confirm provider matches in real time. Rather than framing the solution purely around labor cost reduction, the deployment dramatically scaled outbound contact capacity far beyond what human agent pools could manage, lifting end-to-end appointment completion rates. As Sahil summarized, modern enterprise AI initiatives must be measured against a multi-vector value framework: not merely reducing operating expenses, but driving tangible top-line revenue acceleration.
[14:52](https://www.youtube.com/watch?v=rEGgNLAMocM&t=892s) Best Practices for Scaling Voice AI: Alignment, Maintenance, and Clear ROI
Reflecting on broader enterprise deployments, voice agents consistently deliver across multiple dimensions rather than functioning purely as a cost reduction play. For many enterprise clients, simple cost cutting actually ranks as a lower priority compared to expanding capabilities and unlocking new value vectors. When considering what separates successful implementations from failed initiatives, specific organizational disciplines stand out as critical determinants of victory.
```mermaid
flowchart TD
accTitle: Critical Pillars for Enterprise Voice AI Success
subgraph Alignment["1. Cross-Functional Alignment"]
OPS["Operations & Frontline<br/>(Support, Sales, Outbound)"]
PROD["Product & Tech Teams"]
OPS <--> PROD
end
subgraph Knowledge["2. Ongoing Lifecycle & Knowledge"]
KM["Knowledge Management & SOPs"] --> UP["Policy & Process Updates"]
end
subgraph Metrics["3. Defined ROI Framework"]
REV["Revenue Growth & Conversion"]
EFF["Scale Economies & Cost"]
end
The primary differentiator for winning organizations is deep alignment between operations and the product or technology teams. Projects frequently break down when the operational leaders who manage customer support, inbound inquiries, or outbound sales teams are disconnected from product roadmaps, creating a massive gap in ROI realization. Furthermore, organizations frequently underestimate the continuous "care and feeding" required to sustain voice agents in production. This challenge stems not from flawed underlying technology, but from the fact that business policies, workflows, and products are constantly evolving. Compounding this, many enterprises manage their internal knowledge bases poorly, creating a major initial hurdle where standard operating procedures (SOPs) must first be cleaned and properly structured before voice agents can execute them effectively.
- **Cross-Functional Synergy**: Bridging the operational teams who understand real-world workflows with the engineering and product groups building the solutions.
- **Knowledge Base Hygiene**: Addressing foundational gaps in documented internal knowledge to establish robust, up-to-date agent operating procedures.
- **Measurable Value Drivers**: Explicitly defining and tracking the exact ROI metrics upfront—whether revenue growth, conversion rates, scale economies, or cost reduction—to maintain business confidence and avoid leadership impatience.
---
[16:31](https://www.youtube.com/watch?v=rEGgNLAMocM&t=991s) Q&A: Enterprise Focus vs. Vertical Niche Voice Solutions
Opening the floor to audience questions, an inquiry is raised regarding how Regal AI positions itself relative to domain-specific or niche voice technologies (such as SoundHound) and whether they leverage similar tooling. While specialized vertical startups frequently emerge to serve targeted small and medium-sized business (SMB) niches—such as handling mortgage servicing, scheduling dental appointments, or serving as a virtual receptionist for med spas—their value proposition relies on narrow, out-of-the-box utility with low implementation barriers.
Regal AI differentiates itself by deliberately targeting enterprise-grade requirements rather than SMB point solutions. Enterprise deployments involve massive call volumes, intricate regulatory governance, strict data privacy, and sophisticated security standards that niche SMB platforms are not architected to support. Consequently, Regal operates primarily in heavily regulated, high-scale sectors—including insurance, financial technology, and healthcare—where enterprise infrastructure, scalability, and robust governance models are paramount.
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=rEGgNLAMocM&t=4s) the only place on earth that's hotter than Dallas, Texas right now — Jarrod Johnson
> [04:08](https://www.youtube.com/watch?v=rEGgNLAMocM&t=248s) guilty by association — Jarrod Johnson
> [08:09](https://www.youtube.com/watch?v=rEGgNLAMocM&t=489s) We decided we would go create the AI technology ourselves, and then we forgot we were a services company. — Jarrod Johnson
> [08:09](https://www.youtube.com/watch?v=rEGgNLAMocM&t=489s) voice is coming next quarter — Jarrod Johnson
> [14:52](https://www.youtube.com/watch?v=rEGgNLAMocM&t=892s) care and feeding
## Video description
Sahil Mehta, CCO, Regal AI
Jarrod Johnson, Chief Customer Officer, TaskUs
Abstract
AI Agents are no longer just experimental. Enterprises are deploying them across high-volume, high-stakes interactions. In this fireside chat, leaders from TaskUs and Regal will share how enterprise-grade Voice AI Agents are transforming customer interactions across industries. From complex, high-volume conversations in BPO environments to revenue-driving use cases like lead engagement and speed-to-contact, you’ll get a behind-the-scenes look at real deployments, challenges faced, and the outcomes delivered.
Key Takeaways:
* How AI Agents are being operationalized in real-world, enterprise-grade environments
* The measurable impact on customer experience, lead conversion, and operational efficiency
* Lessons learned from scaling AI Agent programs across different industries
Don’t miss this opportunity to hear what’s working now in AI Agent deployments. Walk away with insights you can apply immediately. Join us and see what’s possible when AI moves from pilot to production.
TaskUs押注语音AI：实时通话投资回报率达聊天机器人的三倍
语音AI的投资回报率是同质化文本机器人的三倍，因为实时语音互动能加速营收转化，而不仅仅是削减人工成本。
TaskUs与Regal AI探讨了企业联络中心向部署自主语音智能体的运营转型，将低延迟语音架构与同质化的文本聊天机器人进行了对比。发言人强调，语音互动带来的转化投资回报率约为文本消息的三倍，同时还能解锁完整的100%互动遥测数据以实现持续的质量保证。他们以一个医疗平台的案例研究阐述了主动外呼语音智能体如何消除预约流失以推动营收增长，而不仅仅是压缩人力成本。最后，他们强调长期的成功扩展需要在运营与技术团队之间进行严格的跨职能协同、持续维护知识库，并建立优于中小企业单点解决方案的稳健企业级治理。
TaskUs 与 Regal 展开合作推进企业级语音 AI 落地，依托深度企业系统集成、上下文实时事件触发和上线前合成模拟三大技术支柱，已累计支撑超过 2.5 亿次通话。 (0:04)
传统联络中心通常仅能人工抽检 5% 至 10% 的通话，而原生语音 AI 平台能全量分析 100% 的交互遥测数据，消除人为执行方差并支持快速 A/B 测试。 (4:08)
相较于面临同质化且容忍多秒延迟的文本聊天机器人，低延迟对话式语音 AI 能传递更高的客户同理心，实现的转化 ROI 约为文本交互的 3 倍。 (8:09)
在医疗健康平台案例中，主动外呼语音 Agent 有效解决了预约流失与排程瓶颈，将价值主张从单纯降低人力成本转向直接拉动顶线收入增长。 (13:28)
企业语音 AI 项目受阻的主要诱因在于一线业务运营与产研团队脱节，同时企业往往严重低估了维持知识库与动态 SOP 持续对齐所需的运维成本。 (14:52)
Reading summary
Enterprise voice AI adoption is shifting focus from basic labor cost reduction toward top-line revenue growth and continuous operational oversight, according to leaders from TaskUs and Regal. While text-based chatbots remain widespread, native low-latency voice agents achieve higher conversion and consistency across complex, regulated sectors when supported by disciplined cross-functional governance.
Universal Telemetry and Pre-Deployment Testing (4:08)
Traditional contact centers rely on manual quality assurance teams that sample only 5% to 10% of calls to evaluate policy adherence and customer sentiment. Jarrod Johnson of TaskUs argued that agentic voice platforms redefine this baseline by capturing and analyzing 100% of interaction telemetry continuously. This complete data ingestion removes human variance across edge cases and allows organizations to systematically run controlled A/B tests and iterate policies across all customer touchpoints.
To ensure operational safety before going live, enterprise platforms also deploy pre-call simulation engines. Sahil Mehta of Regal explained that executing thousands of synthetic conversation simulations allows enterprises to stress-test workflows, pricing changes, and compliance policies against backend enterprise ontologies, validating agent behavior prior to live customer engagement.
Modality Economics: The High-Conversion Voice Premium (8:09)
While the technology industry has historically pushed text chatbots due to lower compute costs, Mehta noted that text interactions normalize 15-second response latencies that would derail a live voice conversation. Furthermore, text agents face rapid commoditization from platforms like Salesforce Agentforce. In contrast, low-latency conversational voice provides stronger customer empathy and converts at roughly three times the return on investment of text channels, justifying its higher per-interaction compute expense.
TaskUs adopted this focus after attempting to build proprietary generative AI models in-house. Recognizing that enterprise clients buy core AI software from dedicated engineering organizations rather than service providers, TaskUs screened approximately 800 funded AI startups. They selected Regal based on its native telephony infrastructure, latency optimization, and enterprise integration capabilities, rejecting vendors that treated voice merely as a prospective roadmap feature.
Revenue Generation Through Automated Outbound Interventions (13:28)
Production deployments demonstrate that voice agents generate substantial business value through revenue acceleration rather than simple headcount reduction. In a deployment for a venture-backed mental wellness platform connecting patients with licensed providers, the business suffered heavy drop-off before initial consultations due to calendar mismatches, last-minute cancellations, and scheduling friction.
By deploying proactive, context-triggered outbound voice agents, the company reached drop-off patients immediately to navigate complex calendars and reschedule appointments in real time. The automated system scaled contact capacity far beyond the limits of human agent pools, resolving conversion bottlenecks and directly expanding top-line revenue.
Organizational Alignment and Enterprise Governance Demands (14:52)
Scaling enterprise voice AI requires sustained operational discipline rather than simple technical integration. Mehta and Johnson emphasized that implementations routinely fail when technical and product teams operate disconnected from frontline operational leaders who manage customer support and outbound sales. Furthermore, organizations frequently underestimate the maintenance required to keep pace with changing policies, compounded by internal knowledge bases and standard operating procedures (SOPs) that must be heavily cleaned and structured before agents can operate accurately.
These architectural and governance demands also dictate market positioning. While niche vertical AI providers serve small and medium-sized businesses with lightweight point solutions, large enterprises in regulated sectors like insurance, fintech, and healthcare require platforms built for massive call volumes, strict data privacy, and complex systems integration.
Enterprise voice AI transitions from an experimental efficiency tool to core infrastructure only when organizations maintain disciplined knowledge hygiene and close collaboration between product and frontline operations. Sustainable success depends on establishing upfront metrics that capture top-line business expansion rather than viewing automation solely through the lens of labor cost displacement.
[0:04] Jarrod Johnson: the only place on earth that's hotter than Dallas, Texas right now
[0:04] Sahil Mehta: Enterprise-Grade Infrastructure & Services
[0:04] Sahil Mehta: Context-Embedded Event Triggers
[0:04] Sahil Mehta: Enterprise Observability & Pre-Deployment Simulation
[4:08] Jarrod Johnson: guilty by association
[8:09] Jarrod Johnson: We decided we would go create the AI technology ourselves, and then we forgot we were a services company.
[8:09] Jarrod Johnson: voice is coming next quarter
[14:52] care and feeding
Original transcript
[0:04] Well, thank you so much for joining us.
[0:05] Uh my name is Sah Mata. I'm chief
[0:07] operating officer at Regal and I'm
[0:08] joined by Jared who's chief customer
[0:10] officer at Taskus. Uh we're really
[0:12] excited to have the opportunity to speak
[0:14] to you all today. Obviously, the fact
[0:15] that you chose to come out here uh in
[0:17] Las Vegas for this convention beans you
[0:19] are thinking about uh complicated
[0:21] applications of this new technology form
[0:23] uh and we've been doing in in in the
[0:25] real world. So, we're excited to share
[0:26] some of our experiences. But thank you
[0:28] Jared so much for joining us. Uh and
[0:30] we're looking forward to hearing from
[0:31] you and your experiences at Tasks. I
[0:33] guess before we jump in and talk about
[0:34] some of what we've experienced. Would
[0:36] love to just give backgrounds on our
[0:37] organizations to the group here. So do
[0:39] you mind telling everyone a little bit
[0:40] about Tas?
[0:41] >> Yeah sure. Um Tascus is a business
[0:43] process outsourcing company for the most
[0:45] part. Um we've been in business a little
[0:47] bit less than some of the legacy players
[0:49] about 15 years. Um we're 60,000
[0:52] employees around the world. We do about
[0:54] 60% of our work is in customer
[0:56] experience. So kind of your classic
[0:58] contact center, call center work. We
[1:00] also do a bunch of trust and safety and
[1:01] fraud, risk and compliance work. That's
[1:03] about call it 20% of our business. And
[1:05] then we have an AI services business
[1:08] which helps fuels fuel AI for big tech
[1:11] companies. Um you can think of that as
[1:13] helping foundation model developers back
[1:15] you know five years ago was more helping
[1:18] deliver machine learning models into
[1:20] production for clients. So basically
[1:21] helping them refine their products. Um
[1:24] it's given us a unique perspective on uh
[1:26] AI and its journey over the past few
[1:28] years. Um I'm the chief customer officer
[1:31] which means I run all the go to market
[1:32] functions, sales, marketing, our product
[1:34] strategy, service line strategy, our
[1:36] consulting organization and we just
[1:38] recently built our Agentic AI consulting
[1:40] practice maybe 18 months ago or so. Um
[1:44] I'm from Dallas, Texas. This is the only
[1:46] place on earth that's hotter than
[1:47] Dallas, Texas right now. Appreciate it.
[1:50] >> Amazing. Thanks so much for that
[1:51] introduction, Jared. And I know uh
[1:53] speaking on behalf of the team at Regal,
[1:55] we're very excited to call Taskus a
[1:57] close partner in some of the work we're
[1:58] doing. Uh to give you all a quick
[2:00] background on Regal, Regal is an AI
[2:02] agent platform. Uh we our founders uh
[2:05] were senior leaders at Angie, the home
[2:08] services company, where they were
[2:10] running a 30,000 person contact center.
[2:12] And one of the things they realized was
[2:13] that you can use voice in really
[2:16] powerful ways to not only uh uh not only
[2:19] help your customers solve what they're
[2:20] trying to do but drive uh more ROI
[2:23] positive business outcomes. And the
[2:25] challenge had always been that it was
[2:26] hard to scale voice. Voice was
[2:27] expensive. And that's where their idea
[2:29] to start Regal came came to be. Uh since
[2:32] then uh Regal has grown pretty
[2:34] tremendously. Uh we've raised over $80
[2:36] million. uh work with have done over 250
[2:39] million phone calls and work with
[2:40] hundreds of companies to deploy AI voice
[2:42] agents across the enterprise. Uh in
[2:45] terms of just three key things to keep
[2:46] in mind about us in terms of what
[2:48] differentiates us from other platforms.
[2:49] You might even meet some of them at this
[2:51] at this conference. The first is we are
[2:53] exclusively focused on the enterprise.
[2:55] What that means is we invest both the
[2:57] technical infrastructure and
[2:59] professional services oftentimes in
[3:00] partnership with with Jared's team uh to
[3:02] ensure that the AI agents we build and
[3:04] deploy don't just execute on the calls
[3:06] correctly but also take into effect the
[3:09] data from your uh from your customer
[3:10] systems and the rules in your ontology
[3:13] that you're using to drive workflows uh
[3:15] in your business. The second thing is uh
[3:18] we've built an engine that allows us to
[3:20] have uh basically embed your agents in
[3:23] context. So what I mean by that is uh
[3:25] the ability to not just say hey this
[3:27] agent can make or take a call but also
[3:30] say hey we have a rulesbased system that
[3:32] says if this lead acts in this way or
[3:34] this customer acts in this way trigger
[3:35] this type of message or this type of
[3:37] agent to make that call and then the
[3:38] last thing I'd call out is
[3:40] enterprisegrade observability and
[3:41] quality whether that's quality that you
[3:43] already use for human agents like
[3:44] conversational intelligence call
[3:46] transcriptions scorecards or things that
[3:48] are only possible with AI agents which
[3:50] are things like running thousands of
[3:51] simulations of conversations before you
[3:53] put them live. So, it's been a really
[3:55] exciting journey for our teams together
[3:57] and we're excited to share a bit more.
[3:58] Um, as COO at Regal, I oversee our sales
[4:01] and CS teams. So, I guess uh, you know,
[4:04] just to jump right into the first
[4:05] question, uh, you know, uh, Jared Task
[4:08] is obviously, you know, a business
[4:09] process outsourcing organization. So, it
[4:11] may surprise some folks to hear how big
[4:13] you all are betting on AI agents and CX.
[4:15] Uh, tell us a bit about your confidence
[4:17] in doing that. Uh, and just the process
[4:19] of coming to that perspective in your
[4:20] organization.
[4:22] Yeah, I mean we've we've built our
[4:24] business on the back of high growth tech
[4:25] companies. Um so our largest uh clients
[4:29] are all kind of Silicon Valley based
[4:30] mobile apps um online businesses, tech
[4:34] companies, etc. So they by definition
[4:36] we're sort of like guilty by association
[4:38] with them and having to be sort of tech
[4:40] forward. What became clear to us over
[4:42] the past few years was that we we built
[4:46] up this business supporting foundation
[4:47] models and we started to see the impact
[4:51] that was going to probably have on
[4:52] commercial applications, consumer
[4:53] applications. And so we said this is
[4:56] going to happen. Um how how do we try to
[4:59] take advantage of that? So we took the
[5:01] the learnings we had by helping build
[5:03] some of the technology and then said how
[5:05] could we pivot those resources put them
[5:07] back into the marketplace to help
[5:09] clients adopt technology as part of
[5:12] their strategy for customer experience
[5:13] their strategy for service and support
[5:15] for their consumers.
[5:17] >> Yeah. Yeah. That's I mean you guys were
[5:18] really sitting on the front lines with
[5:20] your customers as a core provider to the
[5:22] AI uh to the models that ultimately made
[5:24] these AI voice agents possible. Um, in
[5:27] terms of, you know, I want to talk a
[5:29] little bit about quality. I think one of
[5:31] the interesting things that we've
[5:33] observed is that historically, if you
[5:35] look at the customers that or the
[5:36] companies that rank the highest in terms
[5:38] of CX quality and you go interview their
[5:40] CX leaders, they typically say the same
[5:42] things. Hey, we go invest in people. We
[5:45] go uh, you know, we train them up. We
[5:47] make sure that they're the highest
[5:48] quality people. They stay with us for
[5:50] years and years and years. And that's
[5:51] been the path to quality. But now with
[5:53] AI voice agents, you actually have a
[5:55] different path that's marketkedly
[5:56] cheaper in terms of the economic impact
[5:59] uh and also unlocking new capabilities
[6:01] because of your ability to have you know
[6:03] uh agentic actions that let you uh take
[6:05] actions in other workflows or uh rag
[6:07] which allows you to talk to other
[6:09] knowledge bases. So when you think about
[6:11] quality you know what gives you the
[6:12] confidence that AI agents will match the
[6:15] quality that task us is known for and
[6:16] has invested in so heavily. Yeah, I mean
[6:18] in the end um our customers are
[6:20] basically looking for some equation of
[6:22] quality and value, right? Whether
[6:23] they're trying to move work offshore,
[6:25] whether they're trying to um use staff
[6:28] augmentation or labor pools in the
[6:29] United States even to get work done
[6:30] because they need scale, they're looking
[6:32] for some sort of equation of quality and
[6:34] value. When we look at what's happening
[6:37] right now, especially with voice AI, um
[6:40] there is a significant opportunity to
[6:42] actually drive quality up. And the the
[6:45] fundamental reason for that is you're
[6:47] sitting on droves of data that you would
[6:50] typically not be sitting on. I would
[6:51] argue that probably 80 to 90% of my
[6:54] clients today are sampling, you know, 5
[6:58] to 10% of their calls. They're doing
[7:00] real reviews. Um they're getting
[7:02] sentiment analysis and then making
[7:04] policy decisions, um process and
[7:06] optimization decisions, product
[7:08] decisions based on that. And not even
[7:10] just on what the agent is doing, but
[7:12] based on how the product is functioning
[7:13] actually in market based on customer
[7:14] feedback. The difference that we see
[7:16] with the Agentic AI technologies is
[7:19] natively they are sitting on top of all
[7:21] of that data all the time which provides
[7:23] us just droves of information and then
[7:26] the really advanced platforms like you
[7:28] mentioned allow us to actually simulate
[7:30] and sort of AB test and pilot things
[7:32] policy changes process changes
[7:35] technology changes um to drive higher
[7:38] quality results than they're actually
[7:40] able to achieve today. The other
[7:42] advantage with you know the the AI
[7:44] technology is humans don't operate
[7:47] flawlessly at scale right we actually
[7:50] have the potential with some of the AI
[7:52] based technologies to drive higher
[7:54] consistency in quality at scale than we
[7:58] probably ever have been with humans.
[8:00] There's different trade-offs right but
[8:02] if you get the use cases right that the
[8:04] opportunities for quality standards
[8:06] actually go up. Yeah, I think you know
[8:09] one thing that I know about task us is
[8:12] uh you know your teams are truly experts
[8:14] on the processes and quality governance
[8:18] processes of your customers. Uh and I
[8:21] can speak to this firsthand because you
[8:23] guys ran a pretty rigorous process to
[8:25] even choose uh you know which technology
[8:27] providers to partner with. Can you tell
[8:29] us a little bit more about how you guys
[8:30] approached that overall? Yeah, I mean
[8:32] like I said we, you know, we were kind
[8:35] of dabbling in the space. I would say
[8:36] two and a half years ago, 18 months ago.
[8:38] We built up a team of u technologists
[8:40] and then we tried to do what
[8:42] entrepreneurial founders do like mine.
[8:45] We decided we would go create the AI
[8:47] technology ourselves
[8:49] and then we forgot we were a services
[8:51] company, right? So we weren't very good
[8:53] at it. What it turned out we were good
[8:55] at was we could put technology around
[8:56] the agent. like we were really good at
[8:58] like agent assist AI stuff that we could
[9:00] do but customers clients would not buy
[9:03] technology from a non-technology company
[9:05] we didn't have the engineering chops so
[9:07] we decided to go out and pick best of
[9:09] breed third party AI technology specific
[9:11] in the agentic space because that's
[9:12] where we felt like our biggest
[9:14] opportunity was and we I mean you know
[9:17] you looked at all the industry data and
[9:18] we were looking at you know who are all
[9:20] the players getting funding in the
[9:22] customer support AI space there's you
[9:24] know 800 we narrowed that down to like
[9:27] 24 dozen or something. We ran an RFI
[9:30] process and we were based on you know
[9:32] technology for the most part like
[9:34] technology um like fundamentals
[9:37] references. We were really interested in
[9:39] companies who actually had referenceable
[9:41] clients in the space who could tell us
[9:42] stories about how they delivered impact.
[9:46] We actually didn't care about price. We
[9:47] had a feeling that the market was going
[9:49] to bear out price over time, price and
[9:51] value over time. And then the last two
[9:53] things we looked at vision like how much
[9:55] did we feel like the platform had a
[9:57] vision for not just today's
[9:58] functionality but the future and then
[10:02] was it somebody we kind of like to go
[10:04] have a beer with on Friday night. So
[10:05] like what was the partnership angle?
[10:06] Were they really interested in
[10:07] partnering investing in us and training
[10:09] us? Did they really want to do a joint
[10:10] go to market? You those sort of things.
[10:12] I would say that was like the big those
[10:14] were the big factors. And then you know
[10:16] the other question right is so why
[10:18] Regal? So nobody was exclusively focused
[10:22] on voice at the time that we were
[10:24] talking to the market. So this is like
[10:26] Q4 last year. Everybody was saying voice
[10:29] is coming, voice is coming. Every
[10:31] partner we talked to said oh it's on the
[10:32] road map on the road map. Because of the
[10:35] um contact center communication platform
[10:37] that the Regal team had had. We've been
[10:39] working with them since 2019. Um Alex
[10:42] the CEO and I had you know probably
[10:43] started talking uh maybe late 18, but we
[10:45] started working together in 19. And so
[10:48] they understood the voice space better
[10:51] than any other player we felt and they
[10:52] had a better product roadmap. In
[10:54] addition, it felt like the Regal team
[10:57] because of the CCAs background
[11:00] understood the customer use cases, their
[11:02] technology stacks, their telefony plans
[11:04] like and the things that mattered in
[11:06] voice calls that the other partners and
[11:09] let's be honest 90% of the partners we
[11:11] were talking to were in chatbot, right?
[11:13] They were different flavors of
[11:15] textbased, message-based, agentic AI.
[11:18] That was the primary difference, you
[11:20] know, in terms of why we why we wanted
[11:22] to go to market with Regal.
[11:24] >> Yeah. No, and I'm I'm very appreciative
[11:25] that you did choose us, but voice was a
[11:28] very conscious decision from day one
[11:29] with Regal. Uh I think we all kind of
[11:31] collectively agreed or perhaps deluded
[11:34] ourselves into believing that chat was a
[11:36] better experience for our customers. Uh
[11:39] but the reality was it was a better
[11:40] experience perhaps for a cost basis and
[11:43] even now when you build an AI solution
[11:45] uh the bar for for a chatbot is is a lot
[11:47] lower than what it is for an AI voice
[11:49] agent. Uh if an AI voice agent takes 15
[11:52] seconds to respond and give you the
[11:54] right answer that's not a very good
[11:56] experience for the customer. Whereas
[11:57] with a chatbot that would be normal. Um
[11:59] so I think we all have had our share of
[12:01] experiences with frustrating chat bots
[12:03] where you don't know where you left it
[12:04] off. You you don't know if you actually
[12:06] are still talking to someone or not. Um,
[12:08] we think voice is honestly the purest
[12:10] customer interaction and and the way
[12:11] that everyone would choose if they could
[12:13] prefer. So, it's definitely been a
[12:14] tenant for VAS from day one.
[12:15] >> I would just add, you think about this
[12:17] quality value equation, um, a lot of
[12:20] client use cases that are coming to us
[12:21] and I think we've been surprised by the
[12:24] the traction that we've been able to
[12:26] make over the past six to nine months.
[12:27] Um,
[12:29] there are just a ton of domestic calls
[12:31] out there that are that are relatively
[12:34] low value that could be handled in just
[12:36] a smarter way. And I assumed when we got
[12:40] into the Agent AI space that everybody
[12:42] was going to come to me and want their
[12:44] chatbot platform and IVR replacement and
[12:47] self-help, you know, help center
[12:48] frontend replacement and then move that
[12:50] into simple, you know, chat contact
[12:52] types.
[12:54] Yes, there's demand there, but it's also
[12:55] a super fragmented marketplace. There's
[12:57] a bunch of partners like call call your
[12:58] Salesforce agent force and you know
[13:00] other um software players in the
[13:03] messaging space that are trying to bolt
[13:04] on the
[13:07] we were very clearly first to market to
[13:09] talk to our clients about the use cases
[13:12] that we could solve for them with voice
[13:14] AI and we were very specific about the
[13:18] problems we could solve and the ROI on
[13:19] those although the price point for voice
[13:21] is is higher the ROI on them is like
[13:24] three times better Right.
[13:26] >> Yeah. Any any uh uh specific examples
[13:29] you want to jump in? Maybe we can talk a
[13:31] little bit about or something along
[13:32] those lines.
[13:33] >> Yeah. So, and we've got a you must have
[13:35] gotten a PR release. I didn't get that
[13:36] PR release. We have a mental wellness
[13:39] technology platform that we that we've
[13:41] been working with. Um and they've
[13:43] they've got a big So, it's a marketplace
[13:45] basically connecting um patients to uh
[13:48] counselors, right? And you've got this
[13:51] inbound interest that's coming through
[13:53] all their marketing programs. And the
[13:54] key the key is to try to get them
[13:56] actually set up and actually take the
[13:57] first appointment, right? That's a lot
[14:00] harder than you think. Um, and any of
[14:01] you have done like a doctor sign up or
[14:03] whatever, there's a thousand reasons why
[14:04] it doesn't work. You cancel the
[14:05] appointment. Um, you don't like the
[14:07] times that are scheduled available.
[14:09] Maybe the provider actually cancels the
[14:11] appointment. So, this conversion problem
[14:12] is a major revenue detractor for this
[14:14] brand. We were able to show them that
[14:17] the labor they were throwing against it
[14:18] today was a super high cost, but that
[14:21] wasn't even the point. keep that if you
[14:22] want to spend that benchmark us against
[14:25] it. We think we'll give you a more
[14:26] efficient experience, a better
[14:28] conversion rate, and we can give you way
[14:30] more outbound basically appointment
[14:32] setting conversions than you're able to
[14:34] get today because they're limited by a
[14:36] bunch of different factors.
[14:37] >> Yeah. And so I think you know one thing
[14:39] I want to make sure everyone takes away
[14:40] here is it's not just about a cost
[14:43] reduction play but in other words you
[14:45] know if you think about any sort of
[14:47] initiative where you're investing in new
[14:48] technology you want to perform on three
[14:50] vectors make more money spend less money
[14:52] enable new capabilities and with voice
[14:54] agent this you know this customer and
[14:56] and several others we're working with
[14:57] are able to hit hit on all three of
[14:59] those vectors not just on a cost
[15:00] reduction play. If anything the cost
[15:02] reduction I I found has been you know
[15:04] lower priority for a lot of a lot of the
[15:05] customers we talked to. Um so one last
[15:08] thing I want to ask and then we've got
[15:10] about five minutes left so we'll we'll
[15:11] open it up to questions is um just
[15:13] anything you see from you know customers
[15:15] who are more successful in the way they
[15:17] approach voice AI uh and bringing that
[15:20] into their enterprises what are some of
[15:21] the best practices and just things that
[15:23] you see about winners?
[15:24] >> Yeah I mean I would say number one is op
[15:26] um alignment between operations and
[15:28] product. So whatever your technology
[15:30] organization is or your product
[15:31] organization where we see projects fail
[15:33] is when the teams who actually run let's
[15:35] say customer support or sales or
[15:36] outbound today and really understand it
[15:39] when they're not aligned with product
[15:41] what product's trying to do or techto
[15:42] teams like you just get this massive gap
[15:44] in terms of ROI realization. The second
[15:46] is I don't think clients anticipate
[15:48] upfront all the care and feeding that's
[15:50] going to be necessary to manage these
[15:51] systems over time and it's not because
[15:53] the technolog is flawed or high
[15:55] maintenance. It's because policies
[15:57] change, processes change, products
[15:59] change like all the time. And oh by the
[16:02] way, like people don't manage their
[16:03] in-house knowledge very well to begin
[16:05] with, which is one of the biggest
[16:07] hurdles we have to get through on our
[16:08] initial implementations, right? We have
[16:10] to help them refine their knowledge in a
[16:11] way that the agent operating procedures
[16:13] can be documented and then be effective.
[16:15] So those are the those are the two big
[16:16] things. The third one I would say is
[16:18] really understanding what your ROI is to
[16:19] your point like is it revenue growth? Is
[16:22] it conversion rate? Is it you know a
[16:24] cost reduction? Is it scale economy?
[16:25] Like you got to understand that upfront
[16:26] and then measure it, document it and go
[16:28] back otherwise the business gets
[16:30] impatient.
[16:31] >> Yeah. Well um with that wanted to open
[16:34] up the floor for questions from from
[16:35] anyone in the audience. So thanks so
[16:37] much for listening.
[16:38] >> Hi for Regal. You guys are um business
[16:40] process agentic. How would you compare
[16:43] yourself to Soundhound? Would you use
[16:45] their technology?
[16:47] >> Yeah. Um I'm I'm not super familiar with
[16:49] Soundhoud. Are they a TTS or Yeah. So
[16:53] there's uh and I'm sure everyone here
[16:55] can name a couple. There are a number of
[16:57] kind of vertical specific plays that are
[16:58] coming up and you know it's a countless
[17:00] number of customers that are companies
[17:02] that have popped up and said, "Hey, I'm
[17:03] just going to handle mortgage
[17:04] servicesing or I'm just going to handle
[17:06] appointments for dental clinics or hey,
[17:07] I'm going to be the receptionist for a
[17:09] med spa." And in each of those use
[17:10] cases, what we find is that they do have
[17:12] an advantage especially with SMBs
[17:14] because their ability to just build a
[17:16] very specific and unique use case that
[17:18] basically works well and is as low cost
[17:20] hurdle. Uh what we find with regal is
[17:22] you know the way we compare to to
[17:24] companies like in that category is that
[17:26] you know we are specifically focused on
[17:28] several use cases within the enterprise.
[17:30] So when you get into larger uh larger
[17:33] call volumes uh more complex governance
[17:35] governance needs and security needs
[17:37] that's where where we find ourselves. So
[17:39] uh typically we're not competing as much
[17:41] with those kind of vertical plays.
[17:42] Correct. I mean uh you know we work
[17:44] primarily with insurance, fintech, uh
[17:47] healthcare, several other verticals of
[17:49] that like but yeah our our focus is
[17:50] enterprise.
