Pedro Andrade Podcast Transcript
Pedro Andrade joins host Brian Thomas on The Digital Executive Podcast.
Brian Thomas: Welcome to The Digital Executive. Today’s guest is Pedro Andrade. Pedro Andrade is the vice president of AI at Talkdesk, where he oversees a suite of AI-driven products aimed at optimizing contact center operations and enhancing customer experience. Pedro is passionate about the influence of AI and digital technologies in the market, and particularly keen on exploring the potential of generative AI as a source of innovation solutions to disrupt contact center industry.
As a senior executive, Pedro is skilled at influencing decision, supporting change, and presenting an innate talent for problem-solving. His successful experience in advisory, driving transformation, and involvement in several markets prove his collaborative nature, resilience, and adaptability. Well, good afternoon, Pedro. Welcome to the show.
Pedro Andrade: All right. Thank you very much, Ryan. It’s great being here with you.
Brian Thomas: Absolutely, my friend. I appreciate it, and I know you do quite a bit of travel. You’re currently in Portugal. I’m in Kansas City. So I appreciate you navigating time zones and calendars to get here today. So Pedro, if I could, I’m gonna jump into your first question.
You co-founded your first software company at age twenty-three, went on to build in healthcare and fintech, and eventually landed at the intersection of AI and the contact center industry at Talkdesk. What’s the common thread through those chapters, and how did your background as a serial entrepreneur shape how you approach building AI products inside a high-growth company?
Pedro Andrade: Well, that’s quite the, that’s quite the story, right? At the age of twenty-three, this was very long time ago. But the reality is that there are a couple of things that are very similar to the days that we are living in now. The– At the time, we were in the dot-net boom. So I will put it in the other way the, the dot-coms were booming and that’s exactly where what I was working at the time in healthcare and fintech, launching companies that were providing typically s- typical services that were only on-prem systems, and launching them on SaaS models was quite challenging.
The things that are common is that the– what really matters is the speed of execution and the speed of innovating, not only with the technology, but also innovating with the process, with the services that you provide. I think that right now here at Okta, we are living exactly a same equivalent disruption wave.
It’s the AI that finally gets to a point where you can meet the promise, and the promise of a better customer experience that is being promised for so many, many, many years. But finally, you get to a point where you can, you can do, you can do that. You can really make a change on what is being promised in terms of customer experience, better customer experience flow reducing or eliminating the waiting times having the right response at the right time.
And so the pattern is pretty much the same. Moving fast, iterate fast Try new things. Approach those transformation waves 20 years or 30 years ago on the SaaS and the SaaS models, now on the AI transformation and the artificial market. This is– Speed is everything. Speed is going to give you the ca-capacity to try, fail, test again until you get it right.
And right now with AI, I think things are even much harder if you don’t keep up on the pace. Things are moving really fast, much faster than it– that was when I was twenty-three. And that means that the need for having that DNA of experimenting, trying, moving, and getting a move– focus on the customer value, I think that is what dictates success or failure.
And so I’m feeling exactly the same here on Talkdesk. As Talkdesk is moving Talkdesk is, is known maybe historically as a CCaaS company. But reality today, customers look at us as an AI company, because if there is one place where AI can make a significant change or bringing this technology can make a significant change is on contact centers.
So Talkdesk is an AI company that provides communication, customer experience platforms to our to our customers, not just a, a place where a human can pick up a phone.
Brian Thomas: Thank you. Really appreciate that. You packed quite a bit there. Love the backstory back in the dotcom days. Was challenging, obviously, trying to provide services in this SaaS model or move to that model at the time.
And you talked about that, the correlation. We’re in a similar time with this AI boom, and you talked about how you need to iterate fast. Speed is everything. Lots of testing just to get it right, and you, and you have to be constantly impr-improving in that particular space in now– in today’s environment.
But you are working to improve the customer experience, which I love, and you’re leveraging this technology, and I think you’re doing a great job. So thank you. And Pedro, Talkdesk has made s- a significant strategic shift from being a CCaaS provider to leading with customer experience automation.
You’ve described the CXA as a virtuous cycle: discover, build, orchestrate, and measure. That repeats for every use case. What does that mean in practice for a contact center leader trying to figure out where to start?
Pedro Andrade: Well, that’s exactly the point, Brian. The point is that the reason why our, our cycle starts with the discover is that- Part of our job is to help customers understanding where to start.
You want to enter into a transformation, you are– you need to understand where is your starting point. What are the points where you are feeling most of the friction? Where are the points where your customers are feeling the friction? So understanding that based on base data information that you have stored, recordings, and understanding what– how much time people are waiting to ask you for something, what is their emotional state when they are coming to you, understanding which tools you have available, do– giving you a forecast of improvement, that is a fundamental part of our discover phase.
They are going to give you the tools to have data-based decision on what is the next thing that you need to build to improve your customer experience. So you s- you don’t go in blind. You don’t go with what a, a vendor is telling you, “Oh, you need a chatbot.” Maybe you don’t. Maybe there are other ways.
Maybe your friction points in– are in some-som-somewhere else. And so you need to have those data points that are based on the data, based on your own transcription, not on other customer or other company. It needs to be based on your own data, and that is going to give you the forecasts. Those forecast is going to have two factors: the savings that you are going to have or the improvements in terms of experience, depending on the KPI that you are trying to maximize, and the effort to get it.
When you get those data points, and you get the– your roadmap, your digital AI transformation roadmap, and that is a fundamental step for you to make informed decisions and move to the next phase, which is when you start building. And when you start building, you build on your own, or you count with Talkdesk with our photo deployed engineers that can help you to deliver automations using agentic technology to help you to automate.
Basically it’s an equivalent to take– to hire an AI agent that is going to work alongside with your human agent. That’s this part of the building. You put those AI agents and human agents to work in that moment that I call orchestrate. Basically, it’s a combination and the work, a, a, a partnership between human agents and AI agents that is going to split some of that work.
And then you put them to run. The last step is the measure. You need– the same way you evaluate human performance, you also have tools to evaluate what was the AI performance, and in global, you are going to evaluate your hybrid workforce performance. That cycle then repeats. After you implement a few use cases where you have the most of the pains, you repeat the cycle because now things changed, data changed.
Now you are going to review what is your next pain point, what is your next forecast, and then you repeat the loop. This is how we are helping our customers and our customers also finding what– how they should navigate this AI transformation based on data that they, they own– they have, they have not on forecasts, not on consultancy or consultive predictions, but actually given their own information that they have from their own customers and their own experience with their own their own contacts
Brian Thomas: Thank you. Appreciate that. You talked– And I’ll just highlight some things here, Pedro. Discovery is obviously a big piece of this. You need to figure out where the customer’s pain points are, and there’s tools out there that you can do this. You can mine this information to make those data-driven decisions as you talked about.
But forecasting is key, and you’re looking for those savings and improvements and, what’s the e- level of effort to get to that. And then, of course, you need to make sure once it’s imple-implemented, that continual improvement, you’re measuring and looking at what those outcomes are. So I appreciate that.
Pedro, there’s an enormous gap between the promise of agentic AI in the contact center and what’s actually deployed in production. You’ve described these AI agents as systems that can reason, think, and act autonomously. Where are agents delivering real measurable outcomes today, and where is the gap between the demo and the deployment still the widest?
Pedro Andrade: All right. This this is what I struggle the most and where I spend most of my time. I spend most of my time doing what I call dismantling fairy tales about customers that come to me with cons– stories and, and tales of fantastic m- kind of m- between magic and witchcraft possibilities for, for AI.
So part of my job is bring them to the reality, explain what is possible, what was the demo, what is the difficulty or the issues and the trouble or the complexities. What is the reality that AI can do? What are the things that AI cannot do? What is the difference between a real automation or a fabricated nice file containing a prerecorded automated conversation?
There are– There is a massive gap also because of the, the pressure that all those providers that want to take a bit of this of the pie are putting on, on our customers. And so it’s very common, Brian, that you get customers very very, very confused very disbelieving that what, what can be done and We always had a different approach.
I, I am known from being that person that does the anti-pitch. I always start by, “Hold on a second. This is exactly what is possible. This is what is fantasy. Now, let’s look at what is possible, and what is possible is… can do a great benefit for your organization, for your company, and for your customers, for your customer experience.
Now, let’s dig into that.” So it’s… I- I… Not everyone does that, unfortunately, and I am, I’m– I, I believe that this is a great recipe, is starting from understanding and explaining what is possible, what is not. Now, then drill down into what is possible. So to your question, I’m going to give you a couple of examples.
Majority of the thing… majority of the times, people think about AI in the contact center as a chatbot or a voice bot that can put calls away and deflect calls. But that’s not quite true, Ryan. Let me share with you a couple of a couple of examples where AI, and especially on Talkdesk, we always consider AI as an end-to-end automation journey.
So in, in, in a contact center, there is so many different points that you can automate. It’s not just deflecting a call. This is kind of the thing where people get so blinded on the deflection on the deflection story that because y- they immediately can imagine, “Oh, I can s- I, I can fire those agents.
I, I can… I’m going to save a lot of money.” Reality is that it- it’s going… Maybe that’s a good starting point. Maybe you have other, other places where you can do alongside, of course, doing deflection. There are so many things. Let me give you some examples. End-to-end automation actually means being able to take one interaction that can start as an inbound, or, or can start as well as an outbound an email that comes in a social message that comes from one of the social platforms or a simple phone call And you can actually try to go give provide technology to trying to do things like deflection.
I’m going to resolve this problem on the phone. I’m getting this problem solved. The customer is happy. It didn’t touch a human agent. That is kind of the most common goal that people think when they, they, they, they think about AI in the contact center. The reality is that there is much more than that.
I’m going to give another ex– another one example of how one of our customers was using something that is totally unrelated to deflection. The biggest pain was that every, every call that comes in is a potential sales call, so people call into that line to buy something. The biggest pain is that depending on the type of things that they want to buy, and depending on the type of the customer, if it is a, a, a enterprise customer or residential customer the call needs to be redirected to BPOs that were taking care of those calls.
Biggest– First biggest pain is that traditional IVRs were failing big time. Big time means that they were losing around fifty– forty-five percent of the calls that are– were not going through. The, the– people get confused over the IVR menus. They were dropping the calls. And again, now remember, remember that these were sale potential s- calls that will– would lead to a sale.
Now, the second pain point is that the ones that go through, the fifty-five percent, those calls were misclassified. People get so tired of the, the IVR menus that they just press zero, and they go into one of those BPOs. And what happens is that thirty percent of those calls were misrouted. Then the BPO pick up the call, and they say, “Oh, sorry, this is another line.
I’m going to transfer you.” And then it transfers back to the other BPO. The problem for this customer is that it needs to pay for these BPOs, whatever, independently of the resolution. So, if you transfer to BPO A, and that call needs to be transferred, they need to pay to that one and then pay for the second one.
So two dramas in a row. Number one, almost fifty percent of calls don’t go through, so it, it re- implies- Loss of revenue. Second was the ones that are going through, then 30% of them were misrouted. By– We basically replaced all that experience by just simply re- having a conversation. Hello, welcome to Company X, and we are unable to support you.
How may I help you? By replacing totally the incredible IVR tree by just this sentence, then we use AI to understand and ask s- interaction questions– iterative questions to understand what type of customer are we talking about, what type of services are they looking for, and drive and putting them in the, in the right place.
At the end of the project, which was putting in, in a record time, just in six weeks we transformed totally that experience, and we moved from a fifty percent or forty– a fifty-five percent transfer rate actually to eighty-seven percent transfer rate, and that led to an increase an increase of sales conversion of twenty-six percent.
So this is kind of the transformation that we didn’t deflect any call. It was just routing intelligence, routing to the right place with incredible amount of savings, not just on the savings on the rerouting, but also the incredible in-increase in the– in, in, in sales for that specific customer.
This is another one use case, and if you allow me, just quickly going to another use case. It’s totally unrelated with voice. Contact centers are not just to pick up calls. On the contact centers, you also have people that are working on the back office. They are preparing data. We have one customer that Asks is asked to qualify customers or contacts that they are going to be target of of an outbound sales campaign.
Well, if you qualify to certain– with, with certain criterias, you are going to receive a coupon that you can spend on one of their stores. The problem is that to be able to do that, they were using 14 people, human people, full-time, that were taking a list of candidates, and they were doing one by one evaluating manually eligibility criteria across four different systems manually.
With AI, you have an AI agent that was pulling the data, confirming across these different systems about the avail-availability and, well, well, the doing the qualification of that lead. And all in the sudden, those 14 agents were able to be released to do something else, and that work was totally automated.
Each line, Brian, each, each lead took around fifteen minutes per line to do a qualification. And the conversion rate at the end was so poor, below three percent. So take a look at the amount of money that was spending on people and effort to do lead qualification. Then that lead– that, that, that– those contacts that were qualified were sent out and qualif– and being qualified and, and being basically called out to provide the, the coupon, the voucher.
So as I’m– back to my– back to your question at the beginning, some use cases of where AI can help on customer experience, it’s not just about deflecting. It can be assisting, it can be qualifying, it can be routing. So Talkdesk provides all these sets of capabilities out of the box to help our customers to do much more than, well, a bot or a chatbot and helping end-to-end, and what I dis– what I called at the beginning, an end-to-end automation
Brian Thomas: Thank you. A lot to unpack there. Appreciate that. You talked in the beginning, spent most of your time dismantling fairy tales with your customers, right? ‘Cause what’s, what’s really reality as far as the capabilities of AI and what people have a lot of hype, and I’ve seen this being in this space as well.
But it’s important that you explain to them, and that’s what you do, what’s truly possible, then you drill down on those specifics of the capabilities. You did walk through several examples, I appreciate that, highlighting deflection, customer experience, these IVR pain points, cost, et cetera. And again, that’s awesome.
And Pedro, the last question of the day, so if you could briefly share, as agentic AI continues to mature and the boundary between automated and human interaction keeps shifting, where do you see the contact center industry in five years, and what does the winning customer experience organization look like in a world where AI handles the majority of customer interactions end-to-end?
Pedro Andrade: Okay. That’s a great question. Maybe for five years, I’m not sure if I, if everyone or anyone can tell you where things are going in five years, given this, the pace and that things are evolving. But I can, I can for sure tell you where this is go- where this is going to, where this is going to, In this, in this industry, it is expected that this balance between this hybrid workforce is, is going to shift where we are, where we are getting some level of automation into a full automation automation percentage of the, of the conversations get some level of automation into almost, I would say one hundred percent.
When I– Again, I’m not saying that they are going to be fully… all the conversations are going to be automated and deflected. I’m saying that organizations are now getting prepared for this level of end-to-end automation. And I’m, I’m saying that in the future, what I expect and what I predict is that in the future, you are going to have not agents speaking at the phone calls, but you are going to have agents that are monitoring a fleet of AI agents and monitoring and guarantee that they are doing what is expected and in real-time evaluating if they are performing as expecting and delivering to the quality of the experience that is expected.
So we have been seeing this transformation happening in other, in other areas, like in the retail industry, for example, and I’m expecting this to happen at this in, in the contact center as well. So instead of having a human say, “Hello, good morning,” and has– answering questions about opening hours all day long, I’m expecting that human being to be in front of a, of a, of a console, observing, monitoring, and in making an intervention whenever there is the need, not on escalating, but actually in guiding the machine into the right, to the right experience.
This is this is going– this is what the pinnacle of the hybrid workforce is about. In my view, this is– This is… I’m, I’m super excited about this because I’m, I’m seeing these pieces moving right now in that direction. I’m seeing customers that are already reshaping the structure of their teams. For example, on Talkdesk we have something called CXA Operation Center.
Is a set of tools that already integrates some of this vision that I, I just mentioned. And I’m seeing customers already moving people that before they were picking up phone calls, and they are moving them into this CXA Operation Center as managers that organize and prepare and monitor a fleet of machines.
So this guarantees that this blend between humans and machines guarantees that what you have today, the b- in terms of great capabilities, emotional reaction, empathy, we are going to get that as well. You are always have a human behind the machine. You don’t lose that, and you gain all those other capabilities of fast response, accurate response, which is– this is what is expected from a customer service.
And better than that, my last point where I wish that we get there as soon, as soon as possible is the anticipation. The contact center is the end of the line. You call the contact center when you have a problem. What about if we can anticipate the problem? What about if I, as a customer, I can receive a notification on my cell phone about something that is about to happen, or it may, it may eventually happen, and I avoid the call at all?
So that is kind of my end state, avoiding the con-contact center at all. This is what I use to say, I’m here in the contact center industry to kill the contact center. So I think that kind of states my mental model and the model that we think here at Talkdesk.
Brian Thomas: Thank you so much. I really appreciate that. Again, you have so much information here to share with our audience, but just to highlight a couple things here looking to the future, and you said you don’t believe anybody can accurately predict where AI will be in five years, obviously, due to its unprecedented pace right now.
And you feel that we could have nearly close to 100% end-to-end automation. We don’t know where that’s at gonna be, but what you did talk about is the AI agents will be monitoring in real time to ensure that the customer experience is being delivered what is expected. And that human to machine working hand-in-hand is important.
I, I like how you highlighted that with the human in real time guiding the machine to ensure that we are delivering that customer experience 100% every time, and that’s just awesome, and I’m looking forward to what we can do, the potential with AI today. So thank you. And Pedro, it was such a pleasure having you on today, and I look forward to speaking with you real soon.
Pedro Andrade: All right. Brian, it was my pleasure. Thank you so much for inviting me.
Brian Thomas: Bye for now.
Pedro Andrade Podcast Transcript. Listen to the audio on the guest’s Podcast Page.
