Rebecca Greene, CTO and Co-founder of Regal – Interview Series

Rebecca Greene, CTO and Co-founder of Regal, is a technology executive, product leader, and investor with extensive experience building customer-focused digital businesses. Before co-founding Regal in 2020, she served as Chief Product Officer at Handy, where she led the product, design, and data teams and supported the company’s integration with ANGI Homeservices following its acquisition. Earlier in her career, Greene worked in product management at Amazon (AMZN ) and held strategy, consulting, and financial-services roles at Booz & Company, First Manhattan Consulting Group, and Bear Stearns. She is also a Partner at Blue Trail Partners, where she supports founders building high-impact companies.

Regal is an enterprise voice AI platform that enables contact centers to build, deploy, improve, and manage autonomous AI agents for sales, customer support, and operational workflows. Its platform combines voice technology, advanced prompting, and real-time customer data to deliver personalized inbound and outbound conversations, while integrating with existing customer relationship management, contact center, telephony, and database systems. Founded in 2020, the company is designed to help enterprises automate customer interactions without replacing their existing technology infrastructure.

You helped scale Handy from an early-stage startup into a business that ultimately became part of ANGI, giving you firsthand experience building products through hypergrowth. What lessons from that journey convinced you to co-found Regal, and what customer experience problem did you feel simply couldn’t be solved with existing contact center technology?

While we were at ANGI, my co-founder, Alex, and I were running a 3,000-person contact service center that was handling almost 100 million customer interactions every year. At the start, my goal was to get customers to call us as rarely as possible because everyone in tech was saying “voice is dead,” so we played into that anti-phone mindset. FAQs, chatbots, hiding the phone number on the website. We did everything we could to push customers toward self-service.

But customers kept calling anyway. People really want to get on the phone when something is important, urgent, or too complicated to explain in a text box. That’s when we started to take a deeper look into voice as an outbound channel, as data began to show it was incredibly effective. If you want to get a customer to update a credit card or remind them about an appointment, email gets maybe a 2% click-through rate. With phone calls, we could get 20 to 30% of people to answer. That’s 10x better.

The problem was that no existing contact center technology let companies meet that demand affordably. Voice calls cost roughly $10 each to staff, so most companies were incentivized to avoid them, even when customers wanted them. The gap between the high demand for voice, with no cost-effective way to deliver it well, is what convinced my co-founder and me to start Regal in 2020.

Regal has now handled hundreds of millions of AI-powered customer interactions across enterprise organizations. What were the biggest technical challenges in making voice AI reliable enough for production at enterprise scale, and which breakthroughs made widespread adoption possible?

Regal recently crossed 500 million AI-handled calls, which is about 408 years of conversation. We now have 100+ AI voices, 33 languages, and 26 LLMs, and have generated $9 billion in revenue for our customers.

For us and our clients, the biggest challenge was about making sure that agents had the right context whenever they spoke with a customer and gave accurate advice. We work with companies across some of the most highly regulated industries, like healthcare and insurance, where it’s imperative that the AI understands what it can and cannot say. So, we treat our AI like employees: you have to train it, supervise it, and evaluate its performance constantly.

The real unlock was the arrival of LLMs with genuine reasoning and language generation, which made automating complex, multi-turn calls possible for the first time.

Many people still associate AI voice agents with frustrating IVR systems or robotic chatbots. What has fundamentally changed over the past few years that allows modern voice agents to deliver much more natural conversations?

AI voice agents are completely different from the IVR systems of the past. Anyone still using an IVR is very limited, as they’re deterministic. Today, agents truly reason through a conversation the way a person would, using LLMs for language generation and powered by context and data from 500 million previous conversations. Combined with better voice quality, improved latency response time, and more natural conversational logic, most of our clients’ customers today genuinely can’t tell they’re talking to AI, or don’t mind if they can, as long as it produces the output it needs.

As CTO, how do you balance rapid advances in foundation models with the need for stability, security, and predictable behavior for enterprise customers?

This is the core tension I manage as CTO: foundation models move incredibly fast, but our enterprise customers in regulated industries need predictability, security, and consistent behavior.

We’re not a model provider ourselves; we provide an application for building AI agents, and we sit alongside other contact center software. That separation lets us make improvements without destabilizing the guardrails, integrations, and compliance controls our customers depend on. It’s also why we’ve focused on letting any team build AI agents that handle complex workflows in weeks, not months. We recently launched a product called Copilot, which with one conversation, can build a working first draft of a voice agent for teams with their company’s specific business logic, brand, and security already baked in.

Regal’s platform connects AI agents with customer context and business workflows rather than simply answering questions. Why is access to customer data and business systems becoming just as important as advances in language models themselves?

On its own, a language model can hold a conversation, but it can’t actually do anything for your customer. The moment you ask it to check an order, apply a discount, escalate a case, or update a record, the AI agent needs to be plugged into the same systems your human agents use–like their CRM and billing systems.

As the models themselves grow better in quality, that access to real customer data and business systems is what actually separates a great AI-handled experience from a frustrating one.

Your platform serves organizations across industries such as healthcare, insurance, education, and financial services. What have these highly regulated sectors taught you about deploying AI responsibly while maintaining customer trust?

We intentionally focus on highly regulated, higher-complexity industries rather than the easier low-stakes use cases, in part because of my own background in leading product, design, and data, and working with enterprise contact centers.

Trust and compliance really matter in these interactions. Responsible AI deployment is about making sure the agents have reliable guardrails, accurate account-level context, and a clear path to a human when something falls outside their scope. Trust in regulated industries doesn’t come from AI being flawless, it comes from being predictable and fixable.

Many people assume AI will replace human contact center agents. How do you see the relationship evolving between human employees and AI agents over the next five years, and which responsibilities should always remain human-led?

AI isn’t going to take everyone’s jobs as many feared, but it will evolve what the job looks like.

The more important human role over the next few years isn’t the traditional contact center agent–it’s a whole new role. My co-founder and I are calling it the CX Engineer: someone who manages and oversees the AI agents themselves, working backward from the customer journey rather than sitting in individual calls. You can already see this shift happening more broadly, with AWS just putting $1 billion into a Forward Deployed Engineering unit to help customers build with AI, a role that didn’t really exist a year ago.

We can expect a similar shift in CX: as AI agents handle more of the volume, the people who orchestrate, manage, and quality-check those agents end up being essential hires.

Voice AI has improved dramatically, but customers are becoming increasingly aware they’re speaking with AI. What role do transparency and disclosure play in building long-term trust, and should companies always tell customers they’re interacting with an AI agent?

Customers can always ask for a human, and we build that into every deployment. What we’re actually seeing in practice is that most people either can’t tell they’re speaking with AI or don’t mind once they realize it, because the interaction itself matters more to them than who’s on the other end.

There’s also a structural reason this isn’t talked about more: some companies are already running the vast majority of their customer interactions on AI, but aren’t advertising it yet. I think transparency matters most in the moment a customer asks for it, but isn’t needed as a blanket requirement on every interaction.

With AI now capable of handling increasingly complex conversations, what metrics should organizations focus on beyond simple cost reduction when evaluating whether an AI deployment has actually improved the customer experience?

Internally, organizations should pay attention to how much their own team is actually using AI day to day, and treat heavy usage as an overall good sign. But, the bigger metric companies should consider is revenue per employee. If AI is really working, that number should be going up, and that’s what funds growth and better margins. AI doesn’t replace your people, but the people who use it well end up doing a lot more while retaining the same team. They can serve more customers, close more deals, and build more when revenue per employee is the metric you’re measuring against.

Looking ahead, what emerging technologies or capabilities do you believe will define the next generation of enterprise AI agents, and what developments are you personally most excited to help bring to market over the coming years?

Today, voice is already 60-70% of customer interactions, with the rest split across chat, email, and text. As AI makes voice interactions higher-quality and more capable, I expect voice to become the dominant channel and eventually reach 95% of all customer interactions.

What’s going to define the next generation of enterprise AI agents is how proactive and engaged they are when it comes to customer interactions. The novelty of AI in customer experience is wearing off, and everyone knows that it’s becoming more and more capable, but I’m excited to push its boundaries and have it perform even more complex use cases.

Thank you for the great interview, readers who wish to learn more should visit Regal.