Jason Naparellski, Nick Mehta, welcome. How are you doing?
I am doing great. How are you? Good to see you both.
So, should we just jump in, Julia?
So, this is my first moderating that I've done and I'm not very moderate, so I don't know how well I'll do. But I'll just start by introducing myself. So, my name is Jason Naparellski. I am a former big tech exec. So, worked for Google, worked for Amazon, worked for Oracle, and am now recovering from that. And I just started a new consulting company called the Alpha Alpha Initiative. And the entire purpose of this company is to help CEOs understand, grasp and start to use AI. And the reason why it's at CEO level is that that is the person that needs to understand it the most and be able to implement it throughout the business. I think as well as I'm focused on helping what I would call real businesses, not the kind of businesses I've been a part of, construction companies and things like that that are just starting to understand what this has to do with it. So that's what I'm working on. And Nick, your resume is three and a half miles long. You look, you are central casting for a Silicon Valley executive. As a matter of fact, was Silicon Valley written for you or about you? I'm not sure.
Yeah, totally. I was that the cheesiest, most annoying guy in the show. What was that guy's name? I'm not kidding. Yeah. So, yeah. No, totally. I mean, that when you said stereotype of Silicon Valley executives, I was like, that's definitely not a compliment, but no, I'm just kidding.
I'm glad you say that. But anything else you wanted to mention about your background? And I know you're the CEO of Gainsight, which is a bonafide unicorn, I think, at this point. But what anything else you want to mention about yourself in context of this conversation?
No, except I'm just absolutely obsessed with AI in a way that's probably beyond healthy. So I'm happy to talk about any of this stuff.
So we're focused here on customer service roles within AI and specifically around your company does a lot of work with that. So I'll kind of jump in with the first question and just ask you how is AI transforming the day-to-day work of CSMs right now? Given today's technology, I think it'd be interesting to kind of focus on that first and then how is it changing the skills and the types of things you're looking for in CS people right now.
Absolutely. So the first one, we can definitely dive into this more. We're definitely at the center of all this because we're just obsessed with it. Meet every company all the time, but also doing a lot in our own technology stack around helping people use AI in driving value for their customers as well as a better experience for their employees. And one way I think about where it's happening today and where it's coming is a simple 2x2 matrix, right? So one axis is is this AI usage for internal productivity effectiveness, meaning employees using internally, or is it something people are exposing to their customers? And then the second is is it a mature use case or is it more of an emerging, which is what you were alluding to. So just to give you some examples from each of those. So a mature internal use case obviously, write me an email, summarize this document, things like that, right? Those are very straightforward. One that's becoming more and more mature and I would argue the most valuable thing in probably business but also in usage of AI, but also I'd argue even as individuals, is deep research, which is a feature available in OpenAI Pro as well as in Gemini and maybe some others as well. And it's an agent that you'll, instead of like the traditional query response of like a GPT-4 or o3, it's a longer running agent, it'll run for a few hours and you can go and give it an open-ended research task. So for example, one thing, an example of something that one of our CSMs used, he said okay here's a document of our product functionality, here's a document of what the client said their goals are, they actually the client had given us like a very detailed presentation of their goals, and I want you to go on the web, look at all the information about this client, news, etc. I want you to look at these documents, I want you to think about it for a while, come back to me with an integrated document of how we should pitch our value to this client. And it was awesome and it took, you know, it was just like something somebody would have spent all this time and not done nearly as well. So that's an example of an internal, more mature use case.
Hey Nick, can I stop you there for just, I just want to question you on that last point. Do you think that this puts McKinsey out of business? I had lunch with a partner of not McKinsey but another big one of the other big ones and he's like no I don't think it does. I think what it does though is it makes it so the work that analysts would do doesn't need to be nearly as much. That's what an analyst would do, but nobody hires McKinsey to get that report. Then they want to talk to that partner. But now what's interesting is I can imagine a world where there's some challenge to this world which I'll describe where you're like McKinsey or Bain or BCG or whoever has senior people and just way less junior people and the client mainly is interacting with the senior people because all that administrative stuff is done by AI. But there's a big problem which is well then who are the future senior people because all the senior people used to be junior people so that's a different problem.
And I think Brett was alluding to this a little bit and on that internal use case, what's interesting, the more mature internal, the writing of content, the research, etc., what today most people are just using the LLMs themselves, right? Like individuals are just in companies using them and doing awesome things with them. But what's interesting is like all of the products people are using are building them into them. Think about like when you're on a Zoom meeting and now you have like transcription and follow-up tasks and recommendations in the sidebar and catch me up on the meeting and you could before there used to be apps that you would use to do those things and now it's just built into Zoom. So for example, Gainsight now our products have all the email writing, summarization, all that built in. None of that is novel. I mean, God, like that's like so obvious now, but it's just built into your workflow. And so I think some of the internal AI usage is going to eventually, as it matures, gravitate into the products people use already versus the emerging ones where you're seeking out new technologies or you're using the LLMs creatively yourself. So as an example, we bought a company that's called Staircase and it essentially uses LLMs to analyze all your communication with your customers and pull out all this information about like who they are and where there's risk and all this. It's really cool and that was a startup and we bought it and now it's just going to be part of our product so it's just going to be more mature as it goes. And so I think there's a lot of emerging internal use cases that are largely about analyzing data because we all know like LLM's first class job is language, right? Large language model. But they're getting better at analysis. Most of you hopefully tried like I've got a spreadsheet, help me figure out the meaning of this. Or like something that happens all the time to me now is I'm in an internal meeting and somebody's presenting a slide that was so poorly designed and I don't understand the graph at all because the axes aren't labeled and I just take a picture and I just give it to ChatGPT and say what is happening here and it gives me the full analysis, right? And so I think there's these emerging use cases that over time become more mature. So there's a flow, right? Things start out new and novel and November 30th 2022 or whatever, ChatGPT came out and it was super novel to be like finish this sentence and now it's like no that's so easy, right? But it flows in as it matures it gets built into the existing products people have. I think people are going to constantly now be iterating, trying new things and then hoping they merge in. Now then there's the external side where more and more mature use case is I'm going to give my customer some more self-service capability through AI, self-service chat, instead of searching a documentation site just ask a question like you would ChatGPT, stuff like that, right? Gainsight makes some products there and there's other ones out there. And then the emerging customer-facing use case, meaning like you're giving it to your customer, is all the agentic workflows. So like we have our conference in 3 weeks and we have this agentic workflow around automating the entire process of driving a renewal for a company. The renewal's like you have to communicate with your customer and get some information and great use case for an agent. And so we're actually using Microsoft's agent foundry and we built this super awesome agent that does that. But those are emerging, like that agent we have is an alpha. It's not totally mature yet but it's real, you know? And so I think we're going to have that world where there's always emerging technology and it just will flow in over time into the mainstream products but it'll start out as these research projects and that's why you see companies doing so many proofs of concept on these emerging technologies.
Yeah. So, interesting question for you. You've seen these emails. See the recent one from Fiverr. You saw the one from Duolingo. You saw these CEO emails. Have you sent a CEO email out like that yet?
No, it's I think those are great. Like they set the tone and help people understand. I actually shared those emails. I was like, yeah, this is what's happening. I think people just know it. Like I see it in a slightly different way. I'm like, look, I care a lot about your career long term, and it's not just about Gainsight. Long-term what I care about is you guys being employable way past Gainsight. It makes me super happy when somebody gets a great job after working here and the truth is when people can judge whether it's right or wrong, whatever, you're not going to be employable with like literally as a knowledge worker you will be unemployable very soon. And I even have a friend who runs another company, similar sized, it's a few hundred million revenue, friend who similar sized company and he's like when we do an interview for engineers and they have two interviews, one is write some algorithm with you can't use AI and they probably have some way to verify that or they trust them, and then the other one is write an algorithm with AI and if you don't pass both tests you can't get a job. So if you don't know how to use, in the case of engineering, Cursor or GitHub Copilot or whatever, you will not be able to get a job at that company. And I think that's why I tell our employees look I care about it for Gainsight. I want, you know, we're going to be successful if you guys know how to use AI. I care about even more for your future employability. And I don't think it's exaggeration at all. And in some ways, it's a self-fulfilling prophecy. Companies start doing those memos and they freeze headcount and then they're like only the best of the best and then like the people that are the quote-unquote best of the best are the ones who know how to use AI. It's just a fundamental thing, as fundamental as knowing how to use email.
Right. So you also mentioned in the last what you just said there was something I really want to kind of dig on is you are releasing or you've got some alpha stuff that's coming, right? Lots of it I would imagine because that's what AI enables you to start prototypes and things really fast. So one of the problems that's happening in the gaming industry is slop. So all this AI generated slop coming out. So like games after game after game, you know, you could see Steam is just filling up with just slop because people can vibe code whatever they want and throw it up there, right? I really feel like that's happening in AI right now and I think Google is the biggest victim or biggest person that's doing this wrong and that they're releasing stuff every day that is pre-alpha, you know. So we used to, you went to computer science school and you remember it used to be alpha, beta, right? Gold master, release candidate and then we print the thing and now it's pre-pre-alpha, just throw it out there, put an X post and get somebody to say that they just vibe coded the Declaration of Independence or whatever. Do you, I see this as a risk factor at least especially for early adopters, particularly in my world where I'm working with companies that are working in real businesses, is that they're like okay they hear this, they're scared to death and they want to adopt AI as fast as they can because they want to survive and they want to compete but then they do with this half-baked stuff and it kind of hurts their mentality and they're going to take a step back. Right. Right now after the DeepSeek moment, right, there's been a very big exponential path of releases and everything. Now everyone's kind of taking a step back here and looking at how this matters. How would you advise other CEOs that are getting burned on these things at this point? Just stay the course, keep going, keep trying new things, wait until things are more baked before you implement it or what is your advice?
I think it's such a good question and I think it's not an easy one and probably why you have your consulting business helping people think about the nuances of the pros and cons. And I do think that like we're in a world where going too fast has risk and going too slow has risk and they're both significant and so people are really having a hard time. So I think but that's always been the case.
What is different now, Nick? Like it's always been the case that going too fast and too slow are always bad. But what is now?
Speed. It's like the speed and the existential nature and again all of us have biases one way or the other whether they believe this pace of change and all that. I think it's just going to continue to grow at a level that is unimaginable because the thing that's happening is we've never had technology that can make technology and that can invent technology and it will. And it's starting. And so we're going to be in a world where we will truly have that level of, I'm not saying take off and AI takes over humans or something. I'm not talking about that. I'm just saying the level of innovation is going to grow insanely fast. However, I think that part of the brakes and accelerator thing is the nature of your industry. Is it an industry that is very entrenched, not super competitive, and there's a lot of regulatory risk and privacy risk and other legal risk? Don't go too fast, right? Like what's the benefit of going too fast? Like I don't need my PG&E, which is like the power company in California, I don't need them experimenting with using AI to run their power plant, right? We don't need that, right? But if you're in a highly competitive industry, which by the way, interesting enough, all tech companies are, you have to because the other company is going to and you are going to absolutely be beaten out if you're not constantly accelerating. And so I think the thing and what happens is in highly competitive industries there's less risk because the risk is much more you go out of business than it is you get sued. Whereas at JP Morgan or PG&E or Pfizer, the risk is much more you get sued or you get a whatever, worse, than it is you go out of business. Pfizer is not going out of business tomorrow, you know, but honestly like a lot of software companies, a lot of knowledge-oriented companies can go out of business in the next 5 years if they don't move fast. And so I think part of it's the nature now but I think the other thing you said Jason, so important, as individuals you're right some people try these things and are like oh this is not ready for prime time and then it sets them back. And I think that's a different thing which is a human mindset shift which is basically we are in a permanent beta because we are using, I think most people know, non-deterministic technology at least at a level of the way we look at it. Under the covers it's still deterministic but it's like there's no way for us to understand that determinism. They call that computationally irreducible. Like we can't understand how it works. And so therefore, it's always going to be weird and always going to make mistakes. By the way, humans do too. We don't always say that, but it's true. And so we are going to be in a world forever where there's going to be a bleeding edge and it's going to be a weird bleeding edge and it's going to have lots of problems. And I tell our employees if you're in these dynamic industries or if you want to manage your own individual career you should constantly be frustrated by AI. So yesterday two things happened and like within an hour. I was like let me give deep research, which is a feature as you know very well of multiple LLMs, and it lets you go run a research task that's not like an instant response. It's a batch, it'll go around the web and stuff. And we were in a Slack thread about what's the addressable market for some new thing we're building and they were like we should look at Gartner or Forrester and I just said do deep research and figure this out and got us an answer in like 10 minutes, right? And that was magic. But then like later on, an hour later I had to do a presentation for our company, we had a great first quarter, and I went to make a little graphic of all the awesome things that happened and although image generation in ChatGPT is better, text rendering is still really annoying and I was like this is so annoying. But you know what the difference is? The worst, a month from now or two months from now I'm going to try it again, or maybe a week from now. And I think that's a subtle thing sometimes. Historically people would be like oh this isn't right for me, I'm not going to try it. It's like no I'm going to keep trying it. And so if I give people advice it'd be like keep trying it and if you're not frustrated you're not trying it enough. And that's a hard change humans aren't used to.
