Transformative Tech at the Forefront of Customer Interaction
Apr 22, 2024
Evan Kirstel
Prepare to be captivated by Liz Tsai, an MIT math prodigy whose career arc bends from the stringent world of commodity trading right into the heart of innovative customer support with her brainchild, HiOperator. Today's conversation peels back the layers of her journey, revealing how a curiosity for complex problems and a passion for transformative technology led to the groundbreaking integration of AI in customer service, ensuring the human element remains the beating heart of every interaction.
Listen closely as Liz unfolds the future of customer service automation, where machines and humans work in concert to deliver experiences that are not just efficient but emotionally resonant. We'll dive into the ingenious workings of hiQ, Hi Operator's latest product set to redefine real-time quality monitoring, and ponder the seismic shifts in e-commerce and beyond that are poised to reshape our digital lives. This is where the tapestry of tomorrow's tech meets the tenacity of today's entrepreneurs, all through Liz's lens of leadership and the harmonious melding of tech and operations teams.
Prepare to be captivated by Liz Tsai, an MIT math prodigy whose career arc bends from the stringent world of commodity trading right into the heart of innovative customer support with her brainchild, HiOperator. Today's conversation peels back the layers of her journey, revealing how a curiosity for complex problems and a passion for transformative technology led to the groundbreaking integration of AI in customer service, ensuring the human element remains the beating heart of every interaction.
Listen closely as Liz unfolds the future of customer service automation, where machines and humans work in concert to deliver experiences that are not just efficient but emotionally resonant. We'll dive into the ingenious workings of hiQ, Hi Operator's latest product set to redefine real-time quality monitoring, and ponder the seismic shifts in e-commerce and beyond that are poised to reshape our digital lives. This is where the tapestry of tomorrow's tech meets the tenacity of today's entrepreneurs, all through Liz's lens of leadership and the harmonious melding of tech and operations teams.
More at https://linktr.ee/EvanKirstel
Speaker 1:
Hey everybody, happy Friday. We made it through another week and, boy, I'm excited for this chat today around leveraging GI, gen, ai, ai and beyond in developing amazing customer service experiences. Liz, how are you?
Speaker 2:
Happy Friday. Happy Friday, I'm doing well, I'm excited it's Friday and excited to be here.
Speaker 1:
Well, thanks so much for being here and I'm so intrigued by your mission at Hay Operator. But before we dive into all things contact center, customer service, gen AI and beyond I'm really also intrigued by your background personally and professionally, so maybe we can start from sort of the beginning. I understand you started your educational journey at MIT at 15. That must have been both daunting and exciting.
Speaker 2:
I was definitely the nerd of nerds. I grew up, I lettered I don't know if that's a thing where you went to high school, but I lettered in mathematics and, yeah, went to MIT at 15. Was super, super, duper awkward, but if you're going to be awkward somewhere, mit is a pretty good place to do it.
Speaker 1:
I love that. Well, I'm here in the Boston area and you were at the Media Lab, an amazing institution destination. You must miss it. It's quite a group of technologists and deep thinkers there, Absolutely. Then you went into something that's not related to our discussion today, but commodity trading. Tell us about your career and your journey from that point of view.
Speaker 2:
So in general life I'm a big fan of saying yes to interesting things that come along. So, yeah, I absolutely love my time at MIT and in Boston. I sometimes say that I don't think I appreciated how special a place it is until I left. So I've been very jealous that you still get to be in the Boston area. But after school, you know, I had done my undergrad and did a master's at the MIT Media Lab and I figured, you know, I kind of know how research goes. Let's go and try and do something completely different. So a lot of friends were getting jobs in finance. I interviewed for a few finance firms and then ended up getting offered this really interesting job in physical commodities trading. I applied for the role in the New York City area. They offered it to me in Geneva, switzerland, and I was 21 and said, yeah, absolutely, let's go experience something completely new. And that was what brought me into commodities and really, when it comes to physical commodities, a lot of it is logistics and process optimization as much as it is finance.
Speaker 1:
Absolutely. We could spend a lot of time chatting about that, but then you were bitten by the startup bug with a YC-backed startup. That must have been intriguing and quite a journey as well, before founding High Operator quite a journey as well before founding High Operator.
Speaker 2:
Yeah, so I was in Switzerland, ended up in Singapore for a little bit with the same commodities trading company and then one of my best friends from grad school and really mentors he had just gone through YC and he said, liz, you got to come out here, see what YC and Silicon Valley is all about. And basically wrote my resignation email until I came out to work with him and I loved going from a giant finance company to a team of like 10 people hanging out in Soma in San Francisco and I said there's something interesting here, and when you are a naive and somewhat arrogant 26-year-old you go. Well, I should start a startup too. And that was sort of what led me down the journey of what would eventually become High Operator and High Q today.
Speaker 1:
Fantastic and I have lots of questions around that. But originally, with the mission at High Operator, what was the gap in customer support that you identified? You know sort of the problem statement that led to the creation of High Operator you know, sort of the problem statement that led to the creation of High Operator.
Speaker 2:
So we started High Operator in 2016, 2017, which, with generative AI, you might say we are in a year of the chatbot now, but in 2016, 2017, it was sort of the last time people had rediscovered NLP, and so there's a lot of excitement around taking natural language processing and applying it to customer support. I think because of the amount of text data they sort of get in the support space, and so we got on board with that and we said, okay, we believe the vision. We think the vision is a lot of automation in customer support. But what does that look like? Right, generative AI and LLMs they were. Nothing was really prime time, ready for prime time in 2017 in terms of conversational, rightational chatbots. So we said, well, I bet we can do some work automating the back end of customer support processes, keeping the human in the loop for the conversational piece, and when LMs and conversational AI gets good enough, that'll set the groundwork for us being able to swap out the human and swap in conversational AI.
Speaker 1:
Well, it's quite a vision you had so early on in this space and maybe describe how High Operator is incorporating AI into improving CX. What's your particular approach or philosophy, even on AI these days?
Speaker 2:
So there's always vision and then there's reality, right, and that's what we like to keep in mind sometimes when we're thinking about how you can take probably all the cool advances in generative AI but then actually apply it to businesses who are serving customers. So at High Operator, we started by saying, okay, let's automate the back end. And what does the back end mean, right? So when you I don't know email or chat a customer support agent, the front end is the conversational piece. That's the conversation you're having about your broken item or your replacement or where your order is.
Speaker 2:
The back end is then actually everything that agent has to do in order to actually send your replacement or pull your tracking number or issue a refund. It's going to a bunch of different systems looking at processes and actually making things happen before they come back and tell you what a great job they did for you. So we started automating all of that back end and then that's what sort of laid the groundwork of saying, okay, well, this is a structure, sort of the skeleton of what has to happen, and then, with generative AI in 2022 and 2023, we started layering that on top right. So we're a big believer in that generative AI can make your customer support friendlier. It can augment it, but you kind of need that core skeleton of bag and processes your policies of when you can issue a refund, set those up and then layer generative AI on top.
Speaker 1:
A fascinating approach and talk to us a little bit about your tools. You have a whole suite of tools to transform efficiency and effectiveness of customer support. Maybe talk about the product suite, if you will.
Speaker 2:
BPO in a sense. So this is where we work with consumer companies, and they often have customer service agents, maybe in-house or outsourced already. We'll come along and we'll say, hey, give us some of that volume, send it over to us and we will respond to your customers in an empathetic, scalable manner at a more cost-effective price for you. But then behind the scenes, we're using a little bit of human in the loop, but we're doing a lot of process automation and then generative AI, so you can sort of think of it, as you know, a Superman suit for your customer service agent. So when you use the high operator service, yes, there's humans at the core of it, but they're really sort of managing a whole suite of orchestrated automations and LLMs, right?
Speaker 2:
And then yeah, it's been a heck of a journey to sort of see that thesis play out.
Speaker 2:
And then one thing we've learned over the past really year or so as we went and we started increasing the amount of automation that we use in our platform when we service our customers, one thing we started realizing was that traditional quality assurance processes didn't work as well anymore.
Speaker 2:
So in traditional customer support, you have a team of customer service agents, and then for every 30 agents, you might have one QA analyst.
Speaker 2:
It's their job to go and audit and monitor what the agents are doing, and so as we started automating more and more, we started realizing that traditional QA is pretty slow and it's pretty incomplete. Right, you're only able to QA maybe 3% of what your agents do. And what that starts to make you realize is that as you increase the amount of automation, you now need to be able to automatically monitor the quality of your customer service agents as well. And so a lot of our work over the past year has been really finding ways not just to automate customer support but to automate the way that you monitor customer support, and that's been our latest product suite, hiq, which allows you to use automation not for directly providing support, but to in real time figure out what your customers are writing in, about how your human and your automated agents are replying, and then also predict the customer satisfaction and the result of all the service that you're providing.
Speaker 1:
What a fascinating approach and really intrigued by everything you're doing is in real time. So you do real time grading of customer interactions and what does that mean for training staff or for agents and beyond? How does that real time capability kind of work?
Speaker 2:
A ton of exciting things actually, for both humans and robots and beyond. So for human agents, it's giving them that real-time feedback loop. So as a customer service agent, right, you're doing interactions all day, you're making customers happy, but you might not get results. So you might not get QA results from your manager until a week later. And then we've all been consumers. We're all familiar with the customer satisfaction survey, right? You contact customer support and then two days later you get an email that's like thumbs up, thumbs down how is the support Right?
Speaker 2:
And I'm guilty too, I don't fill out a whole lot of those. But when you fill that out, that's how you send that sort of signal right back to the agent and the company to be like that was great or that wasn't great. So with real-time QA, you don't need to wait for that. You can get instant feedback on every interaction, sort of let you know as an agent how you're doing. But then, as a manager, know across my team of agents human and automated agents how are we performing? So that's exciting from a human perspective and then from an automation perspective. You know, if a human agent doesn't know 100 of what's going on, they're humans. They're awesome if you've hired the right humans, they're going to raise their hands and ask for clarification, right? Automated workflows don't tend to do that as much, so you really have the potential for automated workflows to sort of go off the rails a little bit, unless they're monitoring in real time whether they're providing good or bad quality support.
Speaker 1:
Wow, there's a lot to unpack there. That's super exciting, you know. So talk about deployment. I mean we see a lot of legacy support centers, contact centers, call centers, a lot of technical debt and legacy systems and we still experience so many, you know, difficult, challenging, unsatisfying customer interactions. How do we deploy this technology in this real world environment and leverage all the goodness and opportunity you're presenting today?
Speaker 2:
And that's really the thing, right, I think. As a consumer, I always say have empathy and realize that customer support good customer support at scale is so hard for companies to really provide at times, because customer support is really the glue that holds things together. When things don't go as planned, right, if the thing you ordered arrived on time, early and perfectly, you're probably not contacting support. So support is, by definition, having to pull together a lot of different processes and troubleshoot and all of that. But in our process of breaking through, working and partnering with a bunch of really awesome brands at different points in their automation journey, one of the things that we've really started to lean on is to step one understand your objectives, right. You're never using generative AI because for the sake of using generative AI or automation. So it's understanding your business objectives that you're trying to drive. It's a faster first response times, it's a higher CSAT, it's a higher retention. What is that? And then we like to talk a lot about building that map to automation, right? So we think it's a little bit dangerous to be like zero to fully automate, right.
Speaker 2:
Step one is to sort of construct a roadmap and then instrument every step of that process. That's why we're really excited about our tools like Haikyuu, because it does sort of three things right. It sits in front of all of your tickets. It tells you what your customers are writing in about, because if you know what your customers are reaching out about, you know which ones start to make sense to automate. And then you know QA Scout sits there and monitors all the quality, because, as you know what you want to automate and what your customers are writing in about, you want to understand the quality of responses that you're sending out. And then the satisfaction. The super-sat part of this is then also understanding, once you send those responses out, as you tweak and turn that dial on automation, what impact is that having on customer satisfaction? And then, lastly, how do you then connect that data point to things like customer lifetime value and retention?
Speaker 1:
Fantastic. What's been some of the feedback, the initial responses from your clients who you know, many of whom probably have been doing customer service for decades, some of them some story brands out there. What's been the feedback?
Speaker 2:
It's been positive. I think folks are excited about having a way to really monitor and see these things in real time. But then the other bit to keep in place is that we talk a lot about how automation and we think a lot about how automation really needs to work for all of your stakeholders. So if your customer support team are implementing a new tool, it really it should be good for the customer right. It might be better, faster, more accurate support, but it also needs to be good for the brand from a business perspective. But it also needs to be good for the agents right, because agents, at the end of the day, are often the core of a lot of customer support teams, and so you really do need to go in and say is this a tool that's you're not so much about replacing the customer service agent, but is this a tool that's going to maybe automate some more mundane bits so you can elevate what a customer service agent does?
Speaker 1:
Fantastic. When you're talking to a new customer, what are the first one, two, three steps you suggest as you embark on a journey to leverage Gen AI in the contact center with yourselves? I mean, is there a playbook that you put together? Is it vastly different from customer to customer?
Speaker 2:
Yes and yes. The great thing about customer support is every brand, every team really is unique, but our step one is always to understand goals and then run a bit of a diagnostic. And what we mean by diagnostic is go and really understand everything that their customers are writing in about, try and get a lay of the land of these are our contact types. This is where our contact types are making our customers really happy. This is maybe where there's space for improvement, because I think the exciting thing about 2024, and especially living and running customer service in 2024, is that it's kind of the sexy thing.
Speaker 2:
For the first time, engineers want to work on it. People want to be able to leverage generative AI tools. There's a ton of new stuff out there. But then step one is to really understand, of all the cool tools out there, what's the lay of the land, what can we automate and what are the steps and goals there? And then we can go and explore different tools. Sometimes it's high operators, sometimes it's other tools that we can bring in to accomplish those goals. Right, because lots of shiny things in the generative AI world, but which ones do we let loose on our customers?
Speaker 1:
Oh, I love that Very thoughtful approach. Do you care to call out any customers or partners if they let you? I don't want you to have to pick favorite children here, but you're working with lots of companies and different favorite clients, any industries in particular, for example, that are really adopting your technology aggressively.
Speaker 2:
So we're working with a lot of consumer. That's some of your traditional e-commerce, but then also a lot of digital subscriptions and marketplaces. The latter two have been actually really interesting the last few years because marketplaces there tend to be two sides to that. There's a lot of contact volume and then with digital subscriptions there's been a lot of work around. You know retention and how do you create value for that customer and try and help increase lifetime value. But don't make me pick favorites. We'll shout some out afterwards.
Speaker 1:
Okay, awesome Talk a little bit about the team you built. The culture Seems like really dynamic and fun, but how do you think about the leadership style that you you have and the company you're growing?
Speaker 2:
So it's been. Really, we try and say all the things that we talk about, you know, in terms of aligning stakeholder value when it comes to automating for customer support. We try and bring that into high operator as well, because for quite, for quite a while, we really have to have two sides to high operator. Right, we have our tech side, our software, our sales, all of that, but then for quite a while, with the tech enabled services, we also had our operational side right, and so a lot of what we think about is what does that bridge look like? Right, how do we build really a partnership between operations as well as tech and software, where we do have an operational workforce, but their job isn't so much necessarily to be customer support agents. Their job is to in some sense train the workflows and label data and help make that workflow orchestration platform better and faster, right, as they do their roles as well. So for us, it's about understanding that our customers come first, but that also our team members are really important to making that goal a reality.
Speaker 1:
Oh, sounds so fun. So we're talking about so much technology. It sounds like science fiction, but it's here today, which is what's exciting. But what is next? What is your long-term perspective on this strategic AI implementation and technologies you might be incorporating in the next couple three years?
Speaker 2:
We're excited to see this wave of automation increasing in customer support. So I touched a little bit on this earlier, but it really is an exciting time right now because customer service teams are getting so much love from new software tools and new generative AI tools. But then, as the amount of automation increases in customer support, you really have to start to think about what's the infrastructure that you need to build around. That. One is from a training and learning and development viewpoint, right. What does the agent, two years from now, need to be doing every day if they aren't answering? You know where's my order increase every day, right. How do you level that up? What are the tools around that?
Speaker 2:
And then also the tools around sort of training and monitoring your robots, right. What does that look like? As you increase the amount of automation that's customer facing, how do you then sort of QA and monitor and give feedback at the speed of automation, right? So I think a lot of what I'm excited about over the next two, three years is to see the degree of automation increase, but then also see the infrastructure around monitoring and maintaining your customer support automation grow as well. I think there's the possibility for a lot of really cool tools where you can start to manage customer support teams very analytically with a lot of quantitative numbers around it. Maybe the way that marketing is managed today, Wonderful.
Speaker 1:
It's wonderful to see your approach to leveraging data and science and math to bring real business outcomes. So exciting. Anything on your radar worth calling out, any trips or events or travel coming up worth mentioning?
Speaker 2:
No, I will say it's been great getting out this year and seeing folks in person again. I feel like with the pandemic, a lot of support teams scattered a little bit and 2024 has been a great chance to actually go and see people face-to-face again. But what about you? How are things looking? On your end.
Speaker 1:
Yeah, very fun, very busy. I'll be at RSAC in San Francisco and a bunch of vendor events. So, yes, I agree, you can't beat the human-to-human connection, despite all of this amazing AI technology. That's the best thing in this business. Well, thanks so much, liz. Really intriguing to chat. I hope to see you at MIT at some point, maybe a TED Talk or some other. There's so many great events in the Media Lab. I hope you'll come back to your alma mater and at one of the many events there. And thanks everyone for watching. Reach out to High Operator Any questions. They put out some really great content as well. Thanks everyone, thanks Liz.