Insights from Zendesk Relate - Interview with Tom Eggemeier, Shana Simmons and Cristina Fonseca

25:42Correct. And that is what we built in two hours, right? We built the logs. We've been doing that, right? We've built the auditability, the privacy. We've built there's human and just humanity. This is Shana's colleague that's off camera and off like Mike, who's just butting in. We built the humans. But it's fun. We'll keep that in. That's okay. But I think we've built that all. But I think that's table stakes, right? Like, also what I love that we've built is not just this auditability and logs. I think people are talking about, I have a prediction that next year we are not going to be talking about transparency, for example.

26:25When does someone use AI? Because it's going to be embedded into everything. It's already being embedded into everything. But it's the explainability. Yeah. So it's not just the log, who did what, but it is why did they do that? I think it's better than what we can do as some humans. Yes. But I think we're holding kind of the machines to a higher standard. I agree. This is exactly right. But that doesn't make it wrong. Doesn't make it what wrong? It doesn't make it wrong. Like holding the machines to a higher standard. I don't think it's wrong. No. But I think what... But we are holding them to a higher standard.

26:57But we are holding them to a higher standard. Mm-hmm. Because I say to my team, I said, look, now this agent can tell me exactly why the agent did this. Where did they pull the data from? Why did they make this decision? I can't ask. I don't know the answer to that myself sometimes. And so I think that this is up-level in the customer support that everyone is going to be able to provide. And to be able to fix what's in the back end. So that's a tequila doesn't bring that up. Excellent. So you're building all this trust approach or an approach that is about building trust and securing trust and maintaining kind of - I think we're doing it not just on our platform, but recently we launched an AI observability project where we are also able to look at what your third-party providers are doing.

27:52So if there is a problem on your instance using your platform with the third party, we're able to shut that down. So I think it is about the whole ecosystem that we're looking at and thinking about so that it is not just you're using Zendesk. They've built this governance, they've built the transparency, the explainability, the accountability, but how are they thinking about us and all of, this is a web now of interconnectedness with these APIs and all of this. And we're thinking about all of that with our third party integrations. Well, but then I can think about that from a perspective of like, say something goes wrong and so we've all seen it that happens like a something that happens brand does something or there's a break or something whatever then it's always the brand that gets it in the neck and not the you know and not the the vendor because the brand brands out right so given that you're leading on this trust i feel like domain does that mean that you're increasingly having conversations with your clients around how they should be operating and advising them in that respect as well.

28:59I love it that I was able to move some of the mundane work from my team so that they can sit with our customers and have these conversations, but also our customer support team, our implementation teams, our professional services team, our product teams, are now not only having this conversation about how they can set up their Zendesk instance to be more secure, but we're being more prescriptive and transparent with our customers so that it is a better ecosystem for all. Because while it is your brand that will suffer, we will suffer as well if we're building for the customer, right? We can't just be on the stage playing for ourselves.

29:39We have to build for our customers to hit our numbers. I mean, that's what we should be doing. That's what we're all thinking about. That is true. So by extension of that, hypothetically, would you fire a customer? Yeah. We have a content policy. We have an acceptable use policy, and we have. That's not what we want to be in the business for. No, no. But it is important for our brand. But also what we've done is this was coming up last year, and our customers didn't really understand where their security health was. So we built. No one asked us to. We don't get paid any more money for this. we built a security health overview page that shows the customer how they can set up.

30:17So for example, if you're using terrible passwords, right, it is your name plus one, two, three, it is your birthday, right? We have a security health overview page with health data. We also want to make sure that the customers understand the configuration choices that they are making, and we've built that all into that system. And so I think we are, it is a journey, it is a partnership it is not just one or the other but we want to be able that we we want to be able to bring our customers along the way awesome and so one final thing or a couple of final things actually the one final question somebody wants to build trust or go in this trust building kind of journey because it's a it is a thing and it's an amorphous thing and it's always going to change because we're in this period of change and there's rapid change what some advice would you give or would you share with leaders to say, this is what I would recommend that you do if you want to go on that kind of trust journey?

31:14Wow. That's a good question. Can I answer it in three parts? You can answer it as many parts as you like. I'd say we have to start with the people. Okay. Trust, I think it is very easy for me to talk about the technology and to see how we've built the guardrails in it and how we're thinking about the data and how data sanitization practices and how we're following the law, et cetera. But it has to be about the people at the end of the day. Do you want to do business with me? Do I want to do business with you? Do I trust you? And how do you do that? So there's still that human-centric approach to business and how you treat your employees and your customers.

31:54I think also with the employees, the reasons why I say that is that we're going to be building and orchestrating, and we talked about a little bit this before, these autonomous systems that are working, right? What would be best is if you had a product that worked. Some people think, and you talked a little bit about, some people think that it's sometimes the product. And what I have found, even in my own use of software, I can tell you, this is a really silly example when i got here we had a e-billing for outside so league lawyers use the e-billing software and i was ready to fire the vendor i'm like get the vendor on the phone with me this software is terrible i've lost hundreds of thousands of dollars with outside counsel what i realized is the problem was how we configured it yeah right and so it is going to be just as important that we are building the right thing and we're working with our partners to configure it in a way that works for them so that they are getting the outcomes that they're working on.

32:54And last, I'd say it's to start early on. Make sure that everyone is comfortable using AI, using the tools, and that everyone has the agency to build these systems at work because you can't build something if you don't know what good looks like to one of your points that you've made earlier. I think that's fair. And that's great. And second, well, don't need to. it feels very it's very human first I think we have to be but I think the thing is in this maelstrom of sort of technological kind of evolution and development and things sometimes the human beings get lost in the conversation and it's nice to put the human beings back in the centre because you know we have to have so who are we building these things for?

33:39I know and just because we have a technology that's autonomous it's autonomous still for you who are calling to make your return you who's trying the me who's trying to change my flight yeah my mother who got something on her credit card that wasn't supposed to be there at the end of the day we're still building for me you and my mother yeah no no absolutely one final quick sneaky one what question didn't i ask you that i should have asked you gosh What am I most excited about to see this this week? And I'd say it is our launches that we're making in the employee service space. And it is not because it is our biggest segment, right?

34:25Customer support is our sweet spot. And I think the customers figured out that we could do employee service and start to use it. It was a little afterthought. And now we're putting some focus and some engineering and some focus there. And I'm excited about it so that my team can use it. We've already used it now. We've gotten a lot of the... What's important for me is that my team with AI also, they don't feel lost with it. Yeah. And so we were doing extremely manual, menial work in answering these customer questionnaires. We'd get them. They were thrown over to us. It was simple questions like, what is the company's legal entity name, right?

35:05What is your address? and then actually customers started to ask really good questions about AI, data privacy, data use. Either way, a bot does that now. And you have to think about that person's task, that person that was doing that every day to answer that, didn't have much agency in their job. They didn't have much training to think about or didn't have much training of original thought. They didn't have much time to be intellectually curious. Yeah, yeah. And so that is what I'm most excited about, to get that technology in the hands of my work, because I think people are thinking about it as, well, if they can do more, we don't need as many people.

35:46But it is, what can they do differently? And I've had some extremely great examples on my team now where I'd say they're more engaged, they're more empowered. I don't know, Pam will tell me otherwise afterwards. And they're happier employees as a result. But it doesn't mean that there's not this pressure of transforming into an AI. So I'll come back to the thing I said before we started. And one of the things that got me into this sort of space is there was this continual frustration that organizations often get in the way of people doing a good job. That's true. And I think the more that we can think about how do we get out of the way of people doing that?

36:27And if that means that we can end up using the technology to try and automate or take away a lot of the menial stuff, which frees them up to use their agency and their creativity, and the things that they are trained on or the things that they want to do, then we give people a better chance of being the best versions of themselves. I agree. That's what I look forward to. Perfect. Thank you. It's such a pleasure. Thank you so much. Well, that was another great conversation and some really fantastic insights. A couple of things that stood out for me from my conversation with Shana includes how AI exposes process flows and also the three pillars of evolving AI governance.

37:11Really a lot to think about there. But let's not dwell on that. Let's get into my next chat. And this time I speak to Christina Fonseca, who's the VP of product across the Zendesk. so welcome to another installment of the punk cx podcast from that i recorded here at zendesk relate where i was at their time of the recording the time when it's going to be published it will be last week there's a bit of a time to think and i'm here with christina fonseca who is a vp of product at uh zendesk how are you doing welcome to the podcast thank you so much you're very You're welcome. Now, you were on the main stage.

37:49You were talking about lots of different things. And then there was, as ever, with many of these big events, there's always lots of product announcements. I wonder if you'd give me an overview of some of the things that were announced and some of the big meaty things, the things that are going to be like heavy, important. So first of all, thank you so much for having me. like I would say in terms of the biggest announcements and the biggest things we've been working on have been, I think we started with specialized AI agents for CX. And now we've realized that like the, in the future, like the tasks that are done in the CX domain are going to be, some of them are going to be CX specific, some others not really.

38:32So we've evolved our platform to be a network of agents instead of like our own CX agents. So we have our AI agents that are getting better and better. We've launched ES agents, which has been a little bit of a promise from the past years, like getting into employee service. And we've also launched custom agents. Custom agents give people the ability to basically perform any task you might imagine. And is that where kind of agent builder plugs in? Yes, exactly. And then I would say another big thing that our latest acquisition powers is having CX agents that are not tied to the Zendesk platform.

39:12So I can go to Salesforce customers, First S customers, like ServiceNow customers, and they can use our bot, basically. So this is what we call it. And vice versa. Yes. I mean, that was already happening, right? But we couldn't really, like if you wanted to buy our bot, then you would have to buy the entire solution. And we've changed things. So our AI agents are a little more broad. Then, and I think like a hot topic these days, like there's a lot of things on the co-pilot side. And I think a big theme for our product is how can we make customers successful without them having to put a lot of effort on the configuration side?

39:50Right. Okay. It's like self-service, like getting you to just leverage the data you have, the historical data, because customer service, like there's a lot of historical data you can leverage in order to set things up. And that would already give you a lot of value. But maybe one of my favorite things is what we call admin copilots. I think we launched the first AI-adon three and a half years ago or four years ago, which is a long time ago. and adopting AI from the agent side was never the bottleneck. But having admins commit to it, implementing it, understanding what good looks like is still a problem that's not fully figured out.

40:32Because, I mean, like technology is relatively new. There's so many solutions out there. So the people in charge by the vision and by the fact that they need to invest in AI, but then they're like, okay, like, where do I start? Yes. So Admin CallPilot is this engine that basically tells you what is it that you should be doing next in order to optimize your setup. Okay. How is that configured? Because that's a tricky thing. You're like, oh, no, you've got this thing that's going to tell me what to do. But if I don't really understand this new domain that I'm operating in, then who's telling who kind of what to do?

41:05How does that all get set up? Okay. It's a very good question. So basically what we have is like, it's three main pieces. One is regular insights. Okay. Think of them as an analyst that is continuously looking at your data and gets you insights about what happened in the past. Here I detected - Like successful resolutions - Exactly. That correlated - Right here, problem in there, like data - This is good. This is also good. Okay. Then there's, I would say the most powerful thing is our recommendations engine. Okay. that just scans your data, understands what good looks like based on benchmarks from other accounts and says, hey, if you automatically route these tickets to this team, you will save X, Y, Z.

41:56I think Auto Assist could help your agents streamline the resolution for this particular topic. Or, like, I really see the opportunity for you to automate 15 % of your tickets if you set up an AI agent for this particular case. And that's an agent that's learning across your whole platform. Yes. Anonymized customer base data and all that sort of stuff, but just kind of, like, learning. So it learns based on your data, and it uses your data to benchmark where you are against, like, what we believe is good. And what you believe is good is informed by your experience across your space. It's benchmarking across similar customers, similar industries, and so on and so forth.

42:38But ultimately, this thing leverages your account data to identify improvement opportunities. And it can be you have a lot of outdated knowledge that's not helping either agents or bots resolve anything. Like, go review. And then we have a chat type UI, of course, that allows you to act on all of these recommendations and interact with your admin center this way. Because to be honest, the Zendesk product in the last couple of years just became like a lot more complex, right? And it was supposed to be simple. Exactly. Like it's supposed to be simple, but suddenly we added a lot more products, a lot more acquisitions, and customers really struggle to connect all the pieces.

43:21Right now, you can configure the most complex type of workflow in Zendesk just by chatting with our admin assistant. So we have admins that are like, oh, this is my best troubleshooting buddy. I use it every day. I don't go and click buttons anymore. And to me, this is one of the most maybe differentiated products we're launching because it just allows you to make the ecosystem of Zendesk products much better. Okay. And then I already mentioned a little bit the data aspect, which is very important to us. One of our biggest customer complaints has been analytics.

44:08So another acquisition that we've made last year made it possible for us to revamp the entire analytics infrastructure. so datasets are much more flexible. I can customize the datasets. I have amazing dashboard builders and I have an agentic experience on top. Well, the analytics kind of fuels the orchestration anyway because if you don't generate, you've got the data and if you can't get the analytics, you've got the data and that doesn't do the trigger onto the orchestration and so on and so forth. Exactly. So it's a little bit the loop. Like when we talk about the loop, I need the analytics, I need the signals.

44:40Some are human generated, some others are AI generated and then I need quality assurance. I mean, quality assurance was like we made an acquisition two or three years ago, like at a time where QA was manual, right? Like we had like a front point where exactly like humans would review a sample of the requests. Right now, most of our customers using QA, they have like 100 % QA ratings on every interaction just powered by AI. And we believe our customers are going to get on the AI journey if we just give them the tools. Right. So we are making QA available to everyone. So we've launched Quality Score.

45:19Is that going to then, does it become like a menu of pricing that people's kind of like spend goes up depending on kind of what they do or is it all gets packaged up? So the basic functionality we are putting in the suite. Okay. So available to everyone, same co-pilot. Like a couple of months ago, we've launched co-pilot in the suite. So we're giving every single one of our customers a flavor of what Copilot does. And AI agents, we are also removing barriers and just making it resolution-based. So we are maybe going back to our origins and democratizing the best of what CX has to offer, again, including AI.

45:58Okay. And one thing I wanted to ask, because you mentioned about data and analytics, but there was also an announcement around knowledge Copilot. Now, here's the thing. So I know that Zendesk has been playing with knowledge automation and creation for a number of years. It goes back to content cues that Adrian, the Dormant and his team, was playing around with about six years ago, I think. I narrated that when I joined Zendesk many, many years ago. And then you announced a new knowledge graph product last year. and they had this ability to spot gaps and spin up new articles and then also translate them across different languages.

46:42And I wanted to understand that what's the delta between what you announced last year and kind of this new product? Because I couldn't really see the kind of difference because, and I also don't think that there's, you know, knowledge gets enough. Attention. Exactly. And it's like, I think Adrian kind of like said it kind of like when I spoke to him last year. So it's like if knowledge is like the AI is the next industrial revolution, then knowledge is like the coal. Yeah. Because you don't do anything, all this sort of stuff without that sort of knowledge. No, no. Like the data and the contacts is one of the most important.

47:19The unstructured data is the one thing, but you also have this structured kind of knowledge kind of data as other thing as well. So what's the step on the knowledge kind of products? Okay. So very good. I think like, I mean, I don't want to go back to content use because the technology and the tools were very, very different. No, no, that's kind of fine. I know it's a journey, right? Yeah, exactly. But you're right. So last year with a knowledge graph, like what we've committed to do was to connect to a lot, different sources of knowledge. because that has also been a little bit of a gap in our products.

47:52So we can use the knowledge we have and the data we have internally, but also map that with other sources of knowledge customers use to run their operation. So that was the first step. But now, sometimes, the knowledge base, whatever it leaves, the knowledge base is just not up to date. It's not complete. It's not working. and we need to create this improvement loop. And this is what's embedded in Knowledge Copilot, which is like, I flag, okay, this article is being used in a lot of different flows, but this is not leading to resolutions, which is our ultimate goal. Like you should go and review it.

48:33Okay, like the bot is trying to help customers with this particular type of request, but can't find good knowledge in the knowledge base. I'm drafting you the missing piece and I'm asking you to review and approve. so then next time the bot can use it. So basically the knowledge co-pilot looks at the knowledge, like your Zendesk environment leverages, and makes sure those are complete and up-to-date and relevant for the ecosystem. I see. So there's almost like an ongoing real-time sort of assessment, scanning of the knowledge kind of landscape. So it's like, oh, there's that. That's not quite working.

49:10That's not quite working. That's not quite working. And it's built on top of the admin co-pilot infrastructure, which does this for other areas as well. It's like, okay, I detected the knowledge gap. Here's the recommendation on how to fix it. Like I did half of the work for you, but I needed to review. Yeah. So that's basically. So now I wanna change tack a little bit and move away from events and part-time aspects and stuff because I don't know if people will know, but you are one of the founders of TalkDesk. You also founded and were CEO of cleverly.ai that was acquired by Zendesk about four years ago, I think.

49:51It was the first AI acquisition before AI was a thing. Well, and there was a thing before generative AI took over the world. So you've been in this customer technology space for, what, 15 plus years now. and I just wanted to kind of ask you to kind of reflect on kind of like some of the big biggest changes some of the biggest lessons that you've kind of like learned and also just get your perspective on where do you think the industry's kind of going in the next couple of years I know that's a big thing it's a big thing and like it's a it's break it down into pieces it's a very good question and I think like maybe we are in like the third wave of I mean like talk desk was all about about the cloud movement, right?

50:40Because phones were not in the cloud. We did put phones in the cloud. So it was leveraging the clouds to make CX way easier. I think then AI was, I mean, when we look at the entire industry, there's just a lot of manual work happening everywhere. And AI, in theory, solves that. But it's still not that right. And it still takes a lot of effort in order to truly automate all of this manual work. So when we found it cleverly, that was the assumption we wanted to prove, especially making it scalable. Okay. Because implementing AI still takes a lot of manual effort. Right now, a little less, but I believe in a future where the system just learns by itself.

51:31and we don't need as many like hand-holding and human effort. So at the end of the day, I believe we're still trying to solve the same problems as we were 15 years ago, but now have way more powerful tools. And the bottlenecks are still how can I leverage these technologies to build a system that self-improves and that does the work and we are not relying on humans all the time. And by the way, these women also don't want to do it most of the time. Entering a call center is quite an experience because these people have a job that's not very sophisticated. They do it because they need a paycheck at the end of the month.

52:17And there's better ways to do what they're doing. So I fundamentally believe we are much closer to solving the problem, but we need to get better at building the system that will self-improve and that will rely on data to focus on the outcomes so not a solved thing it's no it's like it i mean it's like it's it's sometimes the problem is really um simple you know and sometimes we because i almost think that i always could say to people look like here's kind of the ideal the ideal thing for somebody is like imagine you have a group of people, multidisciplinary kind of group of people sitting on a table like this and there's a phone in the middle of the table and the customer, somebody calls and somebody picks up and goes, hi.

53:06And then they go, yes, I can help you. And then they go, it's for you. And they're just kind of passive kind of right. And there's no gaps or spaces between it. It's like, that's almost like, well, you, but that's not scalable, but that's kind of almost the solution. How do you can like kind of do that? How do you push as much knowledge and experience to the front as you possibly can? And I think that's the problem that we're trying to solve. How do you aggregate all that kind of data and push expertise and things to the front? So it's interesting you say that we're still solving the same problem.

53:38I think that's it, right? Like if you ask me what's going to happen in the next couple of years, we're still going to try to solve this same problem with better tools. but I think the key answer is how do we create the system that can do it smarter in a smarter world. But then if I think about all of that and I think about the human side of things, you do mention people that work in the contact center and maybe it is everybody that's going from like a customer service rep to everybody who is working as an executive or a professional or practitioner. and there's a lot of people there's a general feeling where people are just going i'm running around kind of like feeling oh my god i don't know what's going on it's like you feel like you're in this maelstrom of just stuff and then there's people talking about all these different changes and how things are going to be you won't get replaced by ai but you get replaced by somebody who can use ai and all these kind of tropes get kind of like run out and I'm like you know but for somebody who's sort of in it who's been at the forefront of some of these these innovations has been an entrepreneur has been a leader has bounded and led to their own kind of companies and is I think is thinking about this a lot and involved in lots of different products and innovations now I mean people are confused and many people are confused and many people are a little bit scared a little bit anxious so what would be your best advice to somebody who might be listening to this to say if they were wanted a bit of a steer to think oh if you did this or this or these things is it more likely to help you to succeed or to maybe even future-proof yourself to maintain your relevance because actually ultimately that's all really one do is be relevant and be useful be useful yeah um so what do you think what advice i know it's a big question no i like i think like that's true but also what ai is also doing is democratizing intelligence a little bit so ultimately what i believe we will see is the role of humans elevated because they they also can access more data more tools and do more complex tasks so like to be honest my advice would be like for people to use all of these tools like just don't be afraid just try them out just understand the possibilities because that's how we up level ourselves like and this is happening not only like in the in the compact center area in product in engineering like we are all going through this revolution like how do i upskill myself and how do i learn like what's my job like when i have access to like a lot more data, a lot more intelligence.

56:31So yeah, like just use the tools. I think Shashi kind of like did a, I was on a panel and kind of a really interesting example about how he's talked about what they've done with the product engineering kind of like team or the engineering team, who said like, we want you to all use the coding kind of tools, but you can't go in and change the code yourself. You've got to actually kind of get used to kind of, and I said it was just... All you can do is give instructions. All you can do is give instructions. You can inspect the code, but you can't change it. You can't physically change it. You have to just tell the tools what to do.

57:06And he said that was just such a massive cultural change. But once you got over it, productivity has gone through the roof. Yeah. For sure. Right? But that's... So before the bottleneck used to be how fast you do all of these individual tasks. Now suddenly you're finding bottlenecks in other parts of the system. So your job becomes to solve those bottlenecks. I think, for example, the role of product managers is really changing because if engineers can code way faster, but we need to be accountable for telling them what is it that they should be working on. Because you can go super fast, but if you go in the wrong direction, that's a problem.

57:49So I think this new agentic coding, like using AI tools to do more and more, also comes with making sure people work on the right things, make sure people understand what the other bottlenecks are, and creates new jobs to be done that maybe were a little invisible before. So it's like it has been fascinating. Here's a hope then, extending that out, is that maybe we'll get to a point that companies will actually realize the importance of the real-time data that the contact center is actually collecting, or the service and support department is collecting, because it will give them all of the insights they need about the things they need to work on to improve.

58:30Yeah. And then there's problems in how you structure that data, right? Like making it coherent. right like the bigger your organization the more pockets of people you have generating data so like there are still a million things for the humans to to to to work on i'm not worried but our individual jobs like this pressure change yeah anything else you um you'd like to add i'm it's an exciting revolution like a little scary sometimes but super exciting perfect thank you so much thank you yet another really great conversation i really enjoyed my chat with christina she's got such a lot of experience as a founder an entrepreneur and a product leader and engineer i mean just brilliant um i particularly liked how she describes endest new product announcements the democratization of core ai tools and her future proofing advice for individuals who are feeling slightly at sea with all of this stuff that's going on right now like many of us do because this is a whole new world out there.

59:33Now before I start musing on that, let's bring things to a close. I really hope that you enjoyed the series of conversations, you got a lot out of them, and that you go and check out some of the new announcements that Zendosk have been making. So do tune in again, and thanks very much. Bye-bye.