AI不安をどう乗り越えるか：ThoughtSpot CEO ケタン・カルカニス氏に聞く

The AI Today podcast, powered by the Project Management Institute, cuts through the AI heist to deliver you proven strategies, real-world examples, and actionable insights that drive AI project success. Tune in to learn from PMI hosts and guest experts. The age of AI is transforming industries, creating a need for leaders who can leverage this innovative technology to drive real results. The University of Cincinnati's Online Master of Science in AI Management prepares professionals to lead AI initiatives across teams, projects, and entire organizations. Explore topics like AI governance, generative AI, machine learning design, and more, while customizing your degree with a stackable, career-focused graduate certificate. Build the skills needed to transform workflows and shape strategic business decisions with the University of Cincinnati. You can learn more at online.uc.edu slash AI. And we'd like to thank them for sponsoring today's episode.

Hello and welcome to the AI Today podcast. I'm your host, Kathleen Mulch. And if this is your first time listening to AI Today, then welcome and thanks so much for joining. And if this is not your first time, then welcome back. We are well into season 10 of AI Today, which is crazy to say out loud that this podcast is now a decade old. So if you've been a longtime listener, you know that the AI Today podcast is now part of Project Management Institute, PMI, and our podcast is officially on PMI.org, which is super exciting. And as we continue to integrate into PMI, we will be interviewing and engaging with various thought leaders, just as we've done throughout the past 10 years of the podcast, also AI practitioners and project professionals to highlight both the AI transformation of project management and the project management of AI transformations. I'm really excited to be joined today by Kayton Karkanis, who is the CEO of ThoughtSpot. So welcome and thanks so much for joining me today. Thank you, Kathleen. Ten years?

Ten years. I know. Whoa, that's incredible. Congratulations. Thank you so much. Clearly, you guys are doing something good. You know, it's funny. The podcast started and was supposed to be about AI today, right? Not really backwards leaning and not too forward leaning. And still, ten years later, have not run out of things to say. Nope, nope, nope. You're in a good business. You'll be okay for the next training at least. I think so. All right. So I'd like to start by having you introduce yourself to our listeners. Yeah, no, thank you. Thank you. Thank you for the opportunity. Well, listen, excited to be on the show, Ketan Karkhanis. I'm the CEO of ThoughtSpot, and we are a data and AI platform, and our mission is very simple. We help companies turn data into an enterprise advantage. This is really important. Our mission in life is not to give you better charts. There are enough products out there that can give you charts. What we are hoping to do is drive data to outcomes, and that's what ThoughtSpot has been focused on.

I've got around 1,000 customers. We are a Gartner Magic Quadrant leader, and the best part of the whole thing is our team and our values. We are going to talk about AI a lot and AI today a lot, but the most important thing I want you to take away about ThoughtSpot is our values. We are a values-driven company because I think of all the things, that is the most important thing in the age of AI. And our number one value is that of trust. Second is customer obsession. So we really stay focused. I feel like, you know, five years from now, all the AI is going to look the same. And we really want to be the company and the platform you trust to run your business. So we wake up in the morning and we are passionate about giving you that trusted platform. Prior to that, spent a lot of time at Salesforce. prior to that Cisco Systems. I'm here in the Silicon Valley, Mountain View, so if anybody is in the neighborhood, come by our offices on Castro Street. Would love to host you guys.

Yeah, thank you for that introduction. So many leaders right now feel pressure to demonstrate AI progress quickly, right? For whatever reason, maybe external, internal pressure that they're feeling. So how do you as a leader distinguish healthy urgency from AI driven fear or anxiety that people may be having? I mean, it's so apparent everywhere. It's not just fear. There's formal, there's executive pressure, there's board pressure, there's investor pressure. Like if you're an executive in a company, I don't think so you can have a single investor or a board meeting without somebody asking you, what are you doing with AI? It's just impossible. No, you can't. Right? It's like you just can't. So, and there's, honestly, it's good. I think so it's good. Now, obviously, we want to channelize that intensity into constructive progress for your company. So what I'm seeing happen a lot, though,

is people are doing what I call the lightweight POC approach to things. to kind of get out of these, oh, look, we did this POC and we saved two hours per week. And if you are using AI to kind of just say I saved two hours per week, I think so you are really, really, really going to be missing the point. Number two is a lot of companies out there, you know, they are not helping customers think through AI beyond the technology. a big part of AI in the company for you all, no matter what level you are at in the company, it's all about culture. It's all about change management. It's more importantly, Kathleen, about not what model you are going to use, but how are you going to rewire your operating model? That is critical. It is the process. How are you going to rewire your processes? The LLM, the model is not going to do that. The operating model is you.

you have to rewire your operating model and for that you need help you need playbooks you need the right cultural foundation you need the right security of the governance you need to know what could go wrong and of course you need the technology so the mistake i think so people make is they think hey we got a flawed subscription i have got ai no what you got is a next-gen search engine you know it was funny i was reading this uh more than 80 90 80 85 percent of usage is what what people are calling search and summarize you know most people are using cloud to search do next deep research and summarize so research and summarize search and summarize but the power of ai goes far beyond that i'm sure we can get into it but but i think like this is the focus hotspot has what we are in fact i'll tell you one funny thing when we sell our products along with the product comes an ai outcome manager and this ai outcome manager works with you hand in glove

uh to help you implement the product and implement the change management and the best practices so uh i think so it's uh there's a lot of ground to be covered here though kathleen there's a lot of ground to be covered here. Yeah, and I like that you bring up AI and change management because that goes hand in hand, right? When you bring AI into the organization and, you know, a lot of people say, yes, I have a subscription to XLLM, Claude or ChatGPT or Copilot, and then they think that they're doing AI, right? And it's like, well, no, you've brought an extra tool now into your toolbox, but you still need to teach people how to use it. And you had talked about trust earlier, And I want to come back to that because trust is so incredibly important when it comes to AI, right? Why? Because people need to feel comfortable using the technology. However, trust isn't always easy to earn. And I think that sometimes, right, the media doesn't necessarily help make that any easier. So what creates trust in an AI-enabled organization?

Number one is a culture of transparency. and that transparency then gets transferred to the architecture and the products you use and i'll talk about both right number one let's talk about the technology the biggest builder of trust is transparency it's kind of straightforward right if you can see it you're going to believe it if you're not going to see it you're going to be a victim of what i call emotion commotion and that emotion commotion results in all sorts of negative behaviors in the organization but if you think about it like the biggest thing people need to realize when you use agents is if you're really using an agent an agent is not just doing search and summarize an agent is interpreting making a decision and actually acting on your behalf so interpret decide and act if your agent is not doing that you're not using agents so that's number one then you're fine then you're

fine. Then you're just using like search, which is okay. But if you're really using AI to drive an agentic enterprise you doing this then that means you need to have transparency which means the technology should be able to show you what is the agent doing How did it arrive at the decision Can you roll back that decision Can there be governance on that decision? Is there an arbitrary checkpoint in the decision? Can there be an audit trail on the decision? Can anybody open the hood and inspect it? So there's a lot of transparency, governance capabilities that need to be built in. How do you know the agent decided correctly? Are there built-in evaluations? Who manages the agents? Is it you? Who gets fired if the agent doesn't do something right? So there's a lot of fundamental choices you have to make in the org design. But on the flip side, forget technology for a second, on the flip side, on the same point of transparency, think about it this way. Organizations should be okay to fail in their internal journeys. and this is like the permission to experiment and fail and i use the word permission is critical

in an organization thinking about embracing ai because look not all of it is clearly laid out right now there will be some things which will be like it was okay it didn't harm you know do no harm obviously uh but it didn't turn out to be this rocket ship we thought it would um and i think so this permission because you have to give, I don't want to say safety, but you want to give support. You want to create a supportive environment for your leaders to push for AI. If you don't do that, then you have to do it top down. And then they're just going to do it begrudgingly. And it's going to be a tick mark like an audit. You know, yeah, we obviously have to go through an audit. So we'll go through an audit. Nobody gets up in the morning and says, yay, I'm so excited I have an audit today. Like, nobody does that. Except the auditors, maybe. But I don't know why we went there.

Coming back to AI and transparency. But this is really important that as leaders we are transparent. That, yeah, we're going to try. The next big builder of trust is the ability to connect AI to ROI. because what happens is people say hey we deployed ai and look it did these amazing five things uh but what did it result in oh nothing or maybe something or they give an answer which five people don't agree on then they all start doubting ai and so there's no trust then they're like yeah it did nothing oh no it didn't really increase my leads it didn't really increase this it really did So connecting AI to ROI means you want to be upfront about the outcome you're trying to optimize and thinking about it that way. So I'm not telling you anything which is not common sense, if you really step back and think about it. But it's critical that we do these basics incredibly well.

Otherwise, you're going to hurry up and wait. And that's what most companies are falling into the trap of. They hurried up. They did these POCs. They did these pilots. And boom, it's all kind of stuck right now. So I think those are a couple of things that build trust. Trust is the biggest velocity multiplier for your AI transformation journey. It's not the model. It's not the LLMs. It's not all those things. It's trust. So I would say, and the third thing I would advise people to think about is challenge. Challenge your technology providers. You should be asking them, like, you know, for example, our agent, Sparter, is an analyst. It's a data analyst, so you can just talk to your data in a whole new way. You can say, hey, what was my churn in North America for the last year? You can go do deep analysis. You can do why. You can do analysis. You can do change analysis. You can do a lot. But the biggest question you can ask Sparter is, how did you arrive at your answer?

This is critical. If the product you're using cannot answer, how did you even come here? Can you show me your work? You know, I always ask for, I run my business on Spar. And Monday morning, before my forecast call, I'm looking at Spar and Spar is giving me my data. One of the things I usually ask Spar is, and I used to do it in the beginning. Now I don't do it anymore. In the beginning, I used to be like, show me your work. How did you do this? Because I'm like, wait, why did you say this deal is at risk? Like, it makes no sense to me. Because the sales rep told me it's great. And then, funny thing happened, Kathleen, you know. Now I've stopped asking that question. Now I don't say show me your work. Just like human relationships, right? Like once the trust is built, the trust is built. Until it's lost. And then that's a problem. But it's just like that. trust is the most killer AI feature. Yeah, I mean, 100% agree, because if you don't trust it, you're not going to use it.

So it doesn't matter how good the model is, how good the solution is. If people do not trust it, they will not use it. Right. Yeah. Yeah. So, you know, you talked about ROI, and I know that tokens is a big cost for many organizations and kind of a hot topic button. Yes. And so first we were talking about token maxing. Now suddenly we've pulled back and we're like, oh, now we have to be careful about, you know, when and how we use AI because we blew our budget for the whole year and we still have months left to go. So you've talked about organizations experiencing this idea of token anxiety. So maybe for our listeners who might not be familiar with that, or if they are, we can make sure we're level setting. Can you explain what you mean by that? And why do you think that so many companies are facing this idea of token anxiety right now? I don't know if you've seen this, but I saw certain customers put token consumption meters on the desktop,

like a small window which is open on the right. So you see how much tokens you're using. That's like bizarre. That's like ration. Like, you know, it's like standing in a line for like rations. Like, you know, are you rationing AI? And it makes sense. Yeah, we have to ration AI because we suddenly got a million dollar bill. Somebody's got to pay it. And people have blown past, CFOs have blown past budgets because of this, right? So that has created this anxiety in leaders. It's like they want to do AI, but there's no way to do AI. And they have to monitor cost and manage it like at a per token basis. and that's like created a negative incentive to using AI. So in a weird way, AI utilization is going to get impacted. And a fundamental to that problem is the architecture of the products that are used out here.

I think so first, let me tell you five years from now, we should not be talking about models and tokens. We should not. uh it's like do you really care which dam the water is coming from in your house as long as you have water like do you really measure how many q-sec feet of water is flowing per minute through the canal like all that is so like it just feels bizarre that's what we are doing right now right that's what we are doing right now is we are measuring all these things and and it's not going to be the case a few years from now because products like ours, ThoughtSpot, we have innovated the architecture, but we have also innovated our business model. I don't know, Kathleen, if you know this, but I don't charge my customers based on the tokens they use. It's unlimited tokens with ThoughtSpot. No token anxiety. No token anxiety. Or like, no. Why? Because we charge you based on outcomes and it's funny customers are more than happy to pay us based on outcomes they're like

yeah if i achieve my outcome you pay i'm like right and then you pay more and then you pay more and then they're happy with it but um i think so there are two parts to it number one is ensuring that we are not token counting you know we are we are focused on roi and outcome counting and you should not be token maxing you should be outcome maxing and if that outcome maxing requires you five tokens, two tokens, or zero tokens? What do you care? You just want to go get the outcome. Sometimes I joke, hey, if you can do it with a notepad and a paper, pen, or pencil, sure. Why not? Go do it. Go do it. Don't use AI just because you should use AI. I know. I feel like everybody wants, you know, they get a hammer, and then everything's a nail, right? And everybody thinks AI is the right solution for every single problem, or they make it the right solution. and I always say it's not. Right. No. You also, you know, you brought up a lot of interesting points. We could probably have an entire podcast just about some of that.

And you talked about outcome-based solutions, right? And so don charge for the input charge right Right You charge for the output And I think a lot of consultants are going that way as well Maybe the industry is also driving people towards that This goes back to that original discussion though around change management right This is a fundamental change into how we have done things in the past, and so people need to wrap their head around it. One thing that I've heard, you know, at PMI I get to travel the globe, and everybody, no matter where they are, says that AI is like the wild, wild west right now. And when you think about that analogy, that's because we didn't really know how far West Bird expansion was, and we didn't know what we would encounter as we continued to move west. And the same thing is with AI. So when people are thinking about token anxiety, really it's kind of that, it's fear, it's uncertainty, right? And they don't always have unlimited funds that if you go over by a lot of money one month, somehow you need to pay that bill, right?

It comes due at some point. So, you know, thinking about this idea kind of of the Wild West, how should leaders think about AI governance in a way that builds trust and accountability without creating bureaucracy? I know that sometimes the word governance is such a dirty word. People think that governance can be about restriction. But when done correctly, it's not. So how should leaders be thinking about this? The role of governance should be acceleration, not slowing down. So if you design governance with that mind frame of I'm going to use governance to go faster, then that's a good point. But I'll tell you, it comes down to the technology choices you're making. for example if you just give everybody a cloud license and say go build your agents then you're going to have agents sprawl you're going to have ai sprawl everything's all over the place why because it's not multiplayer ai uh you need multiplayer ai in your organization such that if somebody does something it can so agents and humans are working together

humans and a humans and agents and agents all three edges of the triangle are working together and so with multiplayer AI, that helps you avoid agents fraud. Two, governance can be a dirty word, but honestly, I have seen customers who have implemented governance with the mindset of building trust. So, for example, you don't want to restrict what you can build. You want to encourage creativity and spur innovation. But if you create a single window that all the agents in my company will be created by the central team, because this is how I'm going to do governance, then it's not going to work. Then it's going to slow down. So what you want to do is enable the edge of your organizations to innovate with AI. It's like it's this is going to be a very silly example, but I'll just give it anyways. for some of your listeners if you go back way in time

okay way in time and I'm going to so date myself right now but it's like you know once upon a time when you had to publish a document in your company you had to go to this team called desktop publishing team a DTP team and they would format your document and they would like structure it and then they would print it for you. I know that sounds bizarre. Okay, but that was the case. There was the central team which would create and structure all documents. Then came Office 365 and all that. And now everybody can create that and everybody can do this and everybody can do this. And now you saw what did that did to the office productivity revolution. The same way with AI and agents. Today, AI is limited to the central AI team, and that's why you're seeing these problems happen. But everybody is going to be able to create an AI agent, and we're going to create that same structure. That's multiplayer AI to do it.

So I feel like, of course, there has to be some level of what we call core controls in place. You know, anybody cannot just log in and access the finance data through their agent. So you need the right controls. Or you can't just fire something that's going to consume a billion tokens. There needs to be some controls. That is no different than the normal controls we have in our offices today. You know, you can't just get in and book a private jet every day. Well, you can't just book it anyways. But you know what I'm pointing to? So those controls are always in place, and you need some of those for AI. But at the same point, I feel like the right technology choices, you know, So that's where AgentSpot, our product, is really taken off. AgentSpot is giving people these governed multiplayer AI environment, which gives them the best of both worlds. Innovation at the edge, but at the same time, governed infrastructure from IT and central process teams.

So you guys can try it out. Actually, go to topspot.com slash AgentSpot. And we have a free, it's a freemium. You can use it for free. No time limit. No time limit. you can go use AgentSpot for free. And you don't even need to enter any code or anything. Just it's available. You don't even need to mention my name. It's free. Okay, well, that's good to know. And I think that that is important. You know, I like how you talk about kind of multiplayer agents and really kind of what you need to be thinking about that, right? Because otherwise you do get AgentSpot and it is real. Right. And I think that that also can help feed into some of the anxiety in general that people have. So one thing that I also wanted to touch upon is I like to call this this idea of augmented intelligence, where you're not replacing the human but helping the human do their job better, but also knowing that sometimes now we are going to be working alongside of different AI tools,

maybe different AI agents. And that's not something that we've done necessarily in the past. I know with RPA, robotic process automation, we had that to some extent. But with AI, it really changes the game. So how should leaders rethink work as AI increasingly contributes alongside human employees? I might say something provocative. If you think of work by saying, I'm going to keep everything, all the jobs the same in my company, which is a very noble way of looking at it. You don't want to impact people. I agree. As a CEO, that's what I worry about every day. But then I think so you're missing the point. It's not going to work like that. You know, some amount of change management is going to be required. Some jobs are going to change and should change. And some jobs are not going to be the same Like before, new, hundreds of new jobs are getting created and new roles are getting created.

I'll tell you an example. Again, this, I'm, now, this goes back. This is so big. It's bigger than the industrial revolution of the 1800s. You know, back then when that happened and the first, do you know there used to be like thousands of people weaving clothes because there was no textile industry, right? the spindle was not there the automated spindle machine was not there or whatever that's called um so that's the industrial revolution brought that and thousands of people are no longer making clothes but at the same time in the in that year or when that happened there was no radiologist job there was no dentist job there was no air conditioner mechanic job there was no air flight engine job there were millions of jobs that didn't exist that got created and the economy move forward. So my point is, be open to the idea of shaping your workforce for the future. That's really important. Number two, be transparent about the change it can bring to your organization.

I think some people want to hear that. They don't want to hear that there is no change because they're not going to believe that. They're just not going to believe you. They're just not going to believe you. They want you to say, yes, there will be some change and we are going to navigate that change in the most empathetic manner we could and a supportive manner. Three, I would say start isolating parts of the roles in your company that are purely non-value add. For example, if you have a person whose job it is only to follow up on accounts receivable invoices, I mean, I think so humans could do better. Humans could do better. Maybe you could have that person follow up and build relationships with customers rather than just press a button. So start identifying roles such as those. In various departments, you will find around 20 to 30 percent of roles that can be really accelerated and automated with AI.

Finally, identify, I'll give you an example, like in our own case, our salespeople. they were spending around almost 20 of their time on things they shouldn be which was taking them away from their customers And I like whoa I not reducing my salespeople In fact, we are hiring. I'm hiring 25. We are already at 105 and I'm hiring 25, 30 more salespeople. But I think so augmentation does play a role, but there will be some roles that you should be prepared to completely refactor. And this is what I mean by don't think about the AI model. think about your operating model. Don't think what LLM you are going to use. Think about what culture you are going to bring in your organization. You know, don't think servers. Think customers. And if you do that, you're going to navigate this just fine. In fact, you're going to be on top of it. Yeah, I like how you brought that up.

You know, this idea, right, of those Luddites that you talked about where they were really resistant to change because overnight, night sometimes they feel that their jobs are going away right and I think that some of it is on the employer right to say we're on this journey together and we are not going to overnight replace you but you have to actually be on right so a lot of people say hey this is a train and it's already left the station well make sure you're on the train and that you're going to be on this journey with us and then if you are and you're trying okay we'll meet you where you're at right and help continue you know, to educate you. One thing that I've heard a lot of CEOs talk about, too, is this idea of an AI literate workforce. And everybody goes, well, we want our employees to be AI literate, but they don't actually define what AI literate is. And I think that that's important to be having those conversations, right, and dig one level deeper and say, we're saying we need an AI literate workforce, but here's what we mean by AI literacy. And it's not just that you can, you know, you have access to a Claude account or a ChatGPT

account, but it's that you actually know how to interact with it, and you know how to apply critical thinking skills, and you can use it to help you do your job better, not just you know how to write a prompt. 100%. I think so. You articulated it perfectly, and that's exactly right. I think so. We are going through this period wherein we have to invest in our teams and our people. You know, one thing we did in ThoughtSpot is everybody, whether you are an engineer, a workplace resource person, an executive assistant, an HR person, or you might be in finance, the entire company had to go through AI training. And it was not like, oh, we're going to test you and grade you after that. There was no test. There was no grade. There was no report. There was nothing. It was all about, we see the world moving.

Here's your opportunity to learn new skills and do something more for your career, for your life, for your family. And this is not just something that will help you at work. AI can help you at home. Being AI literate can be so powerful for your personal life. and when we had that conversation with our teams we saw people jump in so my whole point is do ai literacy invest in that two things to be mindful of number one be uniform everybody in the company has to go through it that's a big mistake i'm seeing people do they're like oh only these people will go through the ai program and these people will not they don't need it right now no no no everybody needs AI. Everybody needs AI. Number two is don't create like these tests and all that stuff and certifications. Oh, don't do that. Don't do that. Because that creates an adverse incentive for people that you're forcing me. I think so you will be surprised at how

is there anybody in the world today, Kathleen, that doesn't want or use the internet? well some folks in the world don't have access to the internet but i think that will will see the benefit from it yes and you don't have to convince somebody to do that they you know the world has four billion people on instagram and facebook and whatsapp and all these kind of things so my whole point is people understand the value of technology and they embrace it our job is to create a path for them to do that. And the role of AI literacy is not a program, it's a path. Just create the path for your people. That's it. Yeah, I think what, you know, and I like that messaging, what you're saying is that create the space for them to learn, right? And then give them the opportunities and also help level set everybody so that AI isn't just a technology issue or it isn't just a marketing issue or it isn't just a sales issue. It's an everybody issue. And so if it is an everybody issue, give everybody the same level of access to learning.

So really like that messaging. Exactly. Okay. This has been such an incredible podcast discussion. I know we could go on forever, but I always like to wrap up now. Season 10, I've been asking this for 10 years. I always ask my guests the same question, and no matter how many times I ask this, everybody brings in their own unique perspective, and I get different answers. So as a final note, what do you believe the future of AI is in general and its application to organizations and beyond? Oh, I'll tell you. I think so we are entering the golden period of humanity. The power of AI on the foundation of the progress humanity has made over the last many, many years and centuries, if you may. For the first time, we have the power to create the idea of abundance, where we could actually eliminate diseases, where we could actually impact food production in such a way that nobody sleeps hungry.

I think so we are entering uncharted territories there's so much we don't know about and we will learn but I couldn't be more optimistic about what lies ahead for humanity with AI and we can talk a lot about work but I think so what AI is going to help us do for education for bringing equality in the world what AI is going to help us do with health with stabilizing our supply chains, with our food supply, I mean, with tackling some of the most complicated problems humanity is facing and will continue to face, it's incredible. And at the same time, it's going to bring new dangers, new perils, new issues that we are going to have to tackle. but no matter what I feel like your show is going to get even more exciting every week

every month because AI is truly this is going to be bigger than the industrial revolution it's not just a step function change for humanity it's a many step function change for humanity and of course we humans we grasp things but we also make mistakes and I'm sure we will make our share of mistakes I am certain of that but the amount of good that's going to come out of it is going to be 100 times bigger than everything else I just look forward to the day when no child is sick I do I look forward to that day Yeah, well, I mean, like I said, I always get such unique responses, and so I loved your answer. And there is no right or wrong answer, which is why I love this question so much. So, Caden, I would like to thank you so much for joining me and for such a wonderful discussion today.

Thank you so much. Glad to be here. The age of AI is transforming industries, creating a need for leaders who can leverage this innovative technology to drive real results. The University of Cincinnati's Online Master of Science in AI Management prepares professionals to lead AI initiatives across teams, projects, and entire organizations. Explore topics like AI governance, generative AI, machine learning design, and more, while customizing your degree with a stackable, career-focused graduate certificate. Build the skills needed to transform workflows and shape strategic business decisions with the University of Cincinnati. You can learn more at online.uc.edu slash AI. And we'd like to thank them for sponsoring today's episode. Thanks for listening to this episode of AI Today. Please make sure to rate this podcast on Apple Podcasts, Spotify, or your favorite podcast platform. And if you want to lead both the AI transformation of project management and the project management of AI transformations,

check out resources from the Project Management Institute, including the CPMAI certification. our various e-learning courses and the PMI by AI blog until you join us next time keep innovating