Home > Columns > Executive Interviews

Talkdesk Executive Interview

Crystal Miceli, Vice President, Head of Product and Industry Marketing, Talkdesk


Click on image below to read the ebook

ebook image

Sheri Greenhaus, Managing Partner at CrmXchange, sat down with Crystal Miceli, Vice President, Head of Product & Industry Marketing at Talkdesk, to explore the future of AI in customer experience. In this conversation, Crystal shares insights on how Agentic AI is transforming customer interactions, improving efficiency, and enhancing agent productivity. She explains how Talkdesk's Ascend AI integrates seamlessly with existing systems, empowering businesses to automate tasks while maintaining a human touch.

Sheri Greenhaus: What are some of the biggest announcements Talkdesk has made?

Crystal Miceli: In 2024, we announced the Ascend AI platform, which can sit on top of any contact center system. This was a significant milestone because it moved us beyond being solely a contact center-as-a-service provider and positioned us as a leading AI company. We've seen strong interest, especially from organizations with on-premises systems that want to modernize but are hesitant due to the complexity of such projects. With Ascend AI, we can seamlessly integrate without requiring a massive overhaul.

We have more announcements coming this year. In October, we expanded Ascend AI from Generative AI to Agentic AI. One of the things I love about Talkdesk is how quickly we can bring ideas to life. Unlike companies that have grown through acquisitions—where innovation can be slowed by overlapping technologies—we were built from the ground up as a cloud-native, single-platform solution. This allowed us to introduce Agentic AI ahead of most competitors.

Another area where we specialize is industry-specific AI. Recently, we launched AI agents tailored for the retail and healthcare sectors. These AI solutions are designed to understand the unique language and needs of their respective industries. And there’s more to come.

Sheri Greenhaus: How do you define Generative AI versus Agentic AI?

Crystal Miceli: Generative AI was a game-changer because it allowed for natural language interactions. It could gather and process vast amounts of information to generate intelligent responses in real time. This was incredibly beneficial for customer service agents, as it helped them quickly find answers and assist customers more efficiently.

Agentic AI takes this even further. Instead of just retrieving and displaying information, it can take action on behalf of an agent or customer. For example, rather than just surfacing a relevant knowledge document, Agentic AI can book an appointment or complete a task automatically. It brings a new level of autonomy to AI-driven customer interactions.

What’s truly remarkable about our Agentic AI is that it’s not just a tool for customer interactions—it can also configure itself. Traditionally, setting up a decision tree or an IVR system meant meticulously programming every possible customer response. For example, if a caller pressed "9," the system would take one action, but if they said a keyword in a slightly different way, additional programming would be required to account for that variation.

Generative AI helped streamline this by understanding natural language, reducing the need for rigid scripting. But Agentic AI takes it even further. Instead of manually defining every step, we simply instruct it with high-level directives—like "act as a kind and responsive customer care agent." It then autonomously asks the right questions, determines customer needs, and routes them to the appropriate agent based on the required skills. It even follows up with, “Can I assist you with anything else?”

This approach shifts programming from complex rule-based logic to intuitive, goal-oriented instructions. If you look at our programming interface, you won’t see lines of code telling it to check for specific words and execute predefined actions. Instead, we’re training it with real-world guidance, allowing it to adapt dynamically and provide a more natural, human-like customer experience.

Sheri Greenhaus: You mentioned programming AI to "be kind." How do you ensure AI behaves that way?

Crystal Miceli: AI interprets kindness as being respectful and matching the customer’s tone. At Talkdesk, we can program AI to adapt to customer interactions, including tone and language, even when multiple languages are used in the same conversation.

What sets us apart is our ability to establish guardrails and training mechanisms. We monitor AI interactions—whether they involve live agents or AI-driven responses—and allow human oversight. If a response doesn’t align with expectations, we can adjust and teach the system how to respond correctly, much like training a human agent. Over time, the AI refines its behavior based on these inputs, ensuring that it consistently aligns with our standards for respectful and helpful interactions.

Sheri Greenhaus: How does the system recognize when something is incorrect? How does it learn?

Crystal Miceli: Some ways we do this include our AI powered quality management capability, which enables examination of 100% of interactions, not just the select few a supervisor might be able to observe. We also offer AI mood insights, which goes far beyond traditional satisfaction surveys to identify how a customer went from being frustrated to feeling satisfied (or vice versa) and then helps administrators optimize future interactions by replicating what worked. The training can all be done in the system, including using our AI Trainer to put guardrails & guidance in place for the AI agents. This level of granularity helps refine customer experiences and improve AI decision-making over time.

Sheri Greenhaus: Are customers actively using this technology now? Has it replaced live agents, or are humans handling more complex tasks?

Crystal Miceli: AI isn’t human enough to replace all customer interactions, nor should it. There are cases where a human touch is essential. Our goal is to make human interactions as effective as possible by ensuring agents have the right information at the right time and by automating routine tasks.

For example, if a customer needs to schedule a doctor’s appointment, an AI can handle that without requiring a nurse to intervene. Simple banking transactions or retail purchases can be fully automated. However, for interactions that involve upselling, fraud detection, or sensitive matters like patient care, a company may make the decision that human involvement is crucial. In these cases, AI supports agents rather than replacing them by handling routine tasks and surfacing relevant insights. This frees human agent time to deal with more complex, critical issues.

Sheri Greenhaus: Where do you draw the line? If AI is handling simple transactions, is sentiment analysis overkill since it’s just going to hand off complex cases to a person?

Crystal Miceli: Sentiment analysis and mood insights are valuable even when human agents are involved because they help coach agents on the most effective ways to interact with customers. Each company needs to define its own balance between AI and human interaction.

For instance, a luxury brand may want to provide highly personalized human service to frequent buyers, and a healthcare provider may also prefer human interaction for certain situations. On the other hand, many customers just want quick, hassle-free resolutions. AI allows brands to align their customer experience strategy with their business goals, whether that means providing efficiency, personalization, or a mix of both.

Sheri Greenhaus: What industries are seeing the most adoption of this technology?

Crystal Miceli: We've seen strong adoption across healthcare—both providers and insurers—as well as in banking, credit unions, and retail. Well-known brands like Michael’s and Humann are using AI to enhance their customer experience. Additionally, we work with utility companies, including Las Vegas County Water District, as well as travel and community service organizations.

Sheri Greenhaus: What benefits are these companies seeing?

Crystal Miceli: The benefits vary by industry, but some of the key improvements include:  

  • Lower abandonment rates – Customers no longer get lost in outdated IVR systems or drop calls out of frustration.
  • Higher customer satisfaction – AI-assisted agents provide faster, more accurate responses.
  • Shorter call handling times – AI helps agents resolve issues more efficiently.
  • Operational cost savings – AI streamlines workflows, making it easier to roll out new programs and campaigns.
  • Business growth opportunities – AI surfaces upsell and cross-sell insights and enables proactive outreach.

One of the unique advantages of Talkdesk is that we don’t just market to industries like retail, healthcare, and finance—we develop AI specifically for their needs. That means customers can start using AI with industry-specific workflows already in place, giving them a competitive edge.

For example, in healthcare, AI can reduce costs by ensuring patients are properly prepared for their appointments. If a patient forgets to fast before a procedure, the appointment may have to be rescheduled, wasting the doctor’s time, hospital resources, and the patient’s opportunity for timely care. Our AI helps prevent these types of inefficiencies by proactively engaging with customers and guiding them through critical steps.

Sheri Greenhaus: So, Agentic AI can either assist agents or fully handle an entire transaction on its own?

Crystal Miceli: That’s correct. It’s entirely up to each company how they want to use it.

What often gets overlooked—but is one of my favorite aspects, especially as someone with a programming background—is that the configuration itself is also Agentic. Instead of manually programming complex workflows, you simply tell the system how you want it to behave, and it adapts accordingly. It’s an incredible leap in automation and efficiency.

Sheri Greenhaus: If a company wants to get started with Talkdesk, what’s the next step?

Crystal Miceli: We take a strategic approach to ensure our customers get real value from the start. Our Value Management Office plays a key role in this process. Their job is to identify success metrics early on and map out a clear path to achieving them.

We begin with a Maturity Assessment to understand a company’s desired outcomes and priorities—both short- and long-term. Then, we analyze their current key performance indicators (KPIs) and forecast how those metrics can improve over time.

We also outline expected costs and benefits to help businesses make informed decisions. From there, our professional services team and our extensive partner network step in to help implement and optimize the solution.

Sheri Greenhaus: For companies using the tool, have they seen measurable improvements in customer effort and satisfaction?

Crystal Miceli: Absolutely. We track these improvements through customer feedback and publish an annual report on the results. Additionally, we commissioned a  2024 Forrester Total Economic Impact Study, which highlighted some key findings:  

  • 60% Containment Rate – Companies using Talkdesk AI saw up to 60% of interactions successfully resolved through self-service, reducing the need for live agents.
  • 75% Reduction in Abandoned Calls – With AI handling routine inquiries efficiently, fewer customers hang up in frustration.
  • Higher Customer Satisfaction – Faster resolutions (up to 25% reduced talk time) and smarter interactions lead to better experiences.
  • Lower Agent Attrition – AI continuously assists and trains agents in real-time, reducing stressful interactions and improving job satisfaction, which increases retention, reducing attrition by up to 20%
  • Revenue Growth – AI-driven proactive outreach and more informed customer conversations—whether live or automated—help businesses identify new revenue opportunities.

These insights reinforce how AI can transform customer service, not just by improving efficiency but also by enhancing both the agent and customer experience.

Sheri Greenhaus: Talkdesk AI can integrate with existing systems, right? It can interface with any platform a company has. How does that work? If a company has a homegrown solution or uses software from another vendor, how would they implement it?

Crystal Miceli: That’s right. To clarify, Talkdesk offers CX Cloud, our end-to-end customer experience platform. It provides voice, SMS, and social media interactions, along with workforce management, shift scheduling, and performance management.

However, we also offer Ascend AI, which can sit on top of any third-party system—whether cloud-based or on-premises. We connect through APIs, allowing our AI to listen to conversations just as it would within our own platform. From there, it can provide real-time guidance, recommend the best next steps, surface relevant knowledge, or even automate entire tasks—all while integrating seamlessly with a company's existing tools.

Sheri Greenhaus: With omnichannel interactions, customers might start on social media, then text, and later call when they need further assistance. If a company uses your full solution, does all that data carry through, so by the time the agent receives the call, they have the complete interaction history?

Crystal Miceli: Exactly. Imagine you’re at the airport, and your flight is canceled. You might be messaging the airline on social media, using their app, and simultaneously waiting on hold with an agent—along with thousands of others dealing with the same issue.

With our system, all those interactions are captured in a unified customer record, whether they involve an AI assistant or a live agent. By the time the customer reaches a human agent, that agent has full visibility into the situation and can pick up where the last interaction left off—no need to start from scratch.

Sheri Greenhaus: What if a company is using another vendor’s system? Can that interaction history still be passed along?

Crystal Miceli: It depends on what their system allows us to integrate with via APIs. Even if it’s not a traditional contact center application, we can still connect and surface key insights for agents. For example, we can provide an agent with a customer’s bank balance, recent transactions, or other relevant data—so long as the customer’s system enables that connection.

Sheri Greenhaus: How would you classify Talkdesk as a company?

Crystal Miceli: There are thousands of companies claiming to be AI companies these days, but at our core, Talkdesk is in the customer experience space.

We are an automation platform that leverages multiple types of AI—Agentic AI, Generative AI, and other AI models—all focused on solving customer experience challenges. Our goal is to automate and enhance interactions, making them smoother, more efficient, and ultimately more satisfying for both customers and businesses.

Sheri Greenhaus: So, you still see Talkdesk as part of the customer service/contact center space?

Crystal Miceli: Yes, but we’ve expanded beyond traditional contact center use cases. We now support sales, healthcare, and other industries that require customer engagement solutions. We focus on intelligent automation for customer interactions.

Sheri Greenhaus: Before we wrap up, is there anything else you’d like to leave our audience with?

Crystal Miceli: One of the biggest barriers to AI adoption is security concerns—many businesses worry AI might compromise customer data. I want to emphasize that security is a top priority for us.

We’ve developed technologies such as biometric authentication and advanced identity verification to ensure that sensitive data—like banking information—is only accessed by the right people.

Additionally, we believe in maintaining a human-in-the-loop approach to AI. This means continuously training AI models, eliminating bias, and implementing strict guardrails to ensure AI behaves responsibly.

My key message is this: AI adoption doesn’t have to be all or nothing. You can integrate AI at a pace that feels safe and manageable for your business while maintaining a high level of security and customer trust.