Exploring AI Innovations in Customer Support: An Insightful Discussion at theCUBE and NYSE Studio
Varun Vummadi, co-founder and Chief Executive Officer of Giga, joins John Furrier of SiliconANGLE Media Inc. in an engaging session as part of the NYSE Wired program at theCUBE's New York studio. The discussion delves into the revolutionary progress in generative artificial intelligence, focusing on how it transforms voice and generative content within various enterprises, particularly in customer support.
In this video, Vummadi shares their expertise on how Giga leverages advanced AI technology to automate customer support processes, minimizing wait times and improving customer satisfaction. With the recent success of a $61 million Series A funding round, Giga is making significant strides in the industry with their cutting-edge AI solutions for enterprises such as DoorDash. TheCUBE Research team and video hosts provide an in-depth perspective on how technology and business strategies converge.
Key insights from this discussion highlight the significant role AI plays in optimizing operational efficiencies as companies look to reduce operating expenses and enhance customer interactions. Vummadi explains Giga's pioneering approach to AI deployment, emphasizing the need for rapid implementation and continuous iterative improvements in AI models. They note that the integration of AI not only streamlines current processes but also opens new revenue avenues, showcasing the transformative potential of AI when correctly applied.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: Mixture of Experts Series
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Mixture of Experts Series.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Mixture of Experts Series
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Mixture of Experts Series. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Varun Vummadi, Giga
In this interview from the theCUBE + NYSE Wired: Mixture of Experts series, James White, CTO at CalypsoAI, joins theCUBE’s John Furrier to unpack CalypsoAI’s newly launched Security Index – the first comprehensive safety ranking of major generative AI models. White explains how the weekly updated leaderboard and the CASI (CalypsoAI Security Index) score enable apples-to-apples comparisons that blend quality and security, helping enterprises move beyond POC purgatory and toward ROI. The discussion connects model selection and risk posture to enterprise strategy at the intersection of tech and finance – where governance, vendor constraints and performance/latency considerations shape deployment choices at scale.
White details CalypsoAI’s Red-Team product and three attack lenses: signature attacks, operational attacks (e.g., overwhelming outputs that mimic denial-of-service) and “agentic warfare,” which uses autonomous agents to probe for jailbreaks and prompt-injection gaps. He breaks down CASI’s inputs across severity, complexity, decay of older tactics (like DAN variants) and defensive breaking points, alongside an average performance column so teams can weigh capability vs. security. Highlights include Anthropic models leading the safety pack (with Microsoft among the leaders), Claude 3.5 scoring 96.25, Claude 3.7 trending into the #2 slot with different security trade-offs, DeepSeek-R1 landing mid-table and GPT-3.5 Turbo dropping from the top 12. White also previews a human-in-the-loop Purple-Team approach, and shares guidance for continuous testing in CI/CD, model family choices across cloud stacks and real-world implications for POCs, benchmarks and production hardening.
>> Welcome back everyone, theCUBE. I'm John Furrier, Host. We here in the NYSE Studio of theCUBE. Of course, we have our Palo Alto studio connecting Silicon Valley and Wall Street. Tech and money are coming together. It's part of our mixture of expert series. We have an expert here who's doing amazing gen AI work in voice and generative content. Varun's here, he's the CEO, Co-founder of Giga. Varun, thanks for coming into theCUBE and to our NYSE studio on the East Coast. Part of our NYSE Wired program. Thanks for coming on.
Varun Vummadi
>> Thanks a lot for having me.
John Furrier
>> We were talking before we came on camera about the cool agents that are coming out, obviously avatars, real life, humans, voice, the role of voice. I mean, it's pretty well documented. All you got to do is get on X, get on Reddit, get on the internet. Voice is the killer app for AI. We all use it on text. We use it with OpenAI. We use it for all the tools. But voice is actually a killer app inside enterprises so you're starting to see money making platforms coming in, leveraging the data. So very hot area you're in right now, so congratulations. Explain for the folks what you guys do, then we'll get into some of the deeper dive.
Varun Vummadi
>> Yeah, sure. For a little bit about Giga, we have recently ran a $61 million Series A. And what we do is we build a customer support agents for enterprises like DoorDash, et cetera, using us. And the biggest adoption, as we are seeing, as you were mentioning is on voice, because you eliminate hold times. None of us like being on hold so we are essentially killing it.
John Furrier
>> I saw a stat from the DoorDash guy, I saw a quote you guys have on the website. I know they're a customer. He said that their drivers could do a variety of different things. And one of the comments he said was, "I want to build a solution that gets them an answer as fast as possible."
Varun Vummadi
>> Yes.
John Furrier
>> Huge important value for him because their company, they have independent drivers and delivery. You don't want to be on hold. Someone pays, it didn't go through. I mean, all kinds of shit happens. They got to speed that cycle down.
Varun Vummadi
>> It's not just drivers, none of us want to be on hold as well. There's a stat, on average, a human spends 48 days of their lifetime on hold. It's like you spend 48 days of your life on hold.
John Furrier
>> We've all been there. We're not saying you do your email. "Oh, okay. Where am I?" You're connected an hour later.
Varun Vummadi
>> One of my inspirations was while earlier in move to US, I called IRS. They had me on hold for six hours. Six hours. I don't know how many times you have dialed IRS.
John Furrier
>> And then of course, it's like fifth hour in it can disconnect.
Varun Vummadi
>> Yeah, disconnect. Oh my God. And you need to keep jumping things. They need to keep transferring you, yeah. It's a pretty big motivation for me to give people... You can just let people get what they want in a very shorter amount of time.
John Furrier
>> Yeah. I mean, during COVID, one of the things that jumped out at me with AWS was Connect, which was their service. It was very clear that low hanging fruit was the data that the corpus of data from all the support. But now that generative AI is here, almost all the successful beachhead positions start with use cases that a company could knock down.
Varun Vummadi
>> Yeah.
John Furrier
>> Call center, customer support. They have domain data. They're getting new data so you have all the elements of good AI right there. Talk about where they're at right now. What are some of the things look like for you guys? What are some of the use cases? What's some of the value? Can you share some stories?
Varun Vummadi
>> Yeah, of course. The biggest, you are correct on point. The biggest enterprise adoption is happening on support and coding because they're like the first use cases that people are able to knock down. But I'm seeing a clear picture of companies who want to automate more things, especially financial services. The biggest op expense of these companies, if you look at, are support and compliance. They have a lot of people taking a look at these things. It's essentially like policies. And instead of a human, AI takes a look at the policies and automates the task. And then again, HR comes next, but support and compliance, companies spend billions of dollars a year on this.
John Furrier
>> They can make or break a customer relationship on this, and this is where I think the frustration turns into opportunity because I'm a consumer. I'll go to a better support whether it's an airline or whatever.
Varun Vummadi
>> Yeah.
John Furrier
>> If I can get an answer quicker, I'm happy. All right, so take me through what's going on inside the enterprise, because obviously there's a couple things that we can review that go back last year and a year and a half, two years. RAG has been popular. Search, that's search.
Varun Vummadi
>> Yep.
John Furrier
>> Same discovery. What's the answer? And then customer support. And both are grounded on the fact that these enterprises have data.
Varun Vummadi
>> Yes.
John Furrier
>> And it's domain specific to their world.
Varun Vummadi
>> Yes.
John Furrier
>> And there's databases everywhere, so you got the combination of these siloed databases. You now have a point of support with data. Existing data, knowledge systems, but also new data. So voice is actually fresh data too, which is a training opportunity, a reinforced learning opportunity. What's your take on all that?
Varun Vummadi
>> Yeah, you're on point. The way when we start working with an enterprise, the way we do it is we consume all the existing data, which sometimes is not even properly documented or written. We consume all the existing data and build this support policies. We keep iteratively improving on it by learning from their existing customer support agents. That is how we do reinforcement learning to keep reading improvement those policies. It's not just where we land. Let's say we ran at 60% resolution rate or 70% CSAT. We want to show companies a clear pathway to reach from 98% or 99%. The only way we can do it is iteratively learning from the humans on what we don't have and what they have.
John Furrier
>> Okay. So talk about the competition. Obviously, this is a highly competitive environment.
Varun Vummadi
>> Yes.
John Furrier
>> Everyone sees this as the low hanging fruit.
Varun Vummadi
>> Yeah.
John Furrier
>> Good use case. What are you guys doing that's different? Can you share the secret sauce?
Varun Vummadi
>> Of course. The big differentiator is a lot of our competitors take Palantir-like approach. It's became very common in AI, which is a forward deployed engineer. We give you engineers. We sit with you in a company and we do it. While it's a great approach, it takes a lot of time to see results. It roughly takes four to six months for them to deploy. We generally go live in less than two weeks. This is, I'm talking with some of the massive enterprises in the world.
John Furrier
>> Two weeks.
Varun Vummadi
>> Two weeks. And we are a product company. We don't have already deployed engineers. As you were mentioning, we take all the data and build the policies and we tell you the agent, you can chat with it and we change it and it can go live.
John Furrier
>> You're basically doing some heavy lifting on behalf of the customer.
Varun Vummadi
>> Yes. We're basically automating the forward deployed engineer path. That is why we can go faster. That's one big thing. The other big thing is iterative improvements. No matter where you land, let's say if you land at 40% resolution rate, because you don't have all the data to resolve the issues at the start, it will iteratively get you to 98% as fast as possible. We also have this exciting launch coming up in a week where we help customers without APIs also to get the benefits faster.
John Furrier
>> All right, so now the agents are coming.
Varun Vummadi
>> Yeah.
John Furrier
>> Where are agents going to fit into your plan? How do you see the agentic piece? Because now that you have some of these capabilities, it's more than a chatbot. There's past completion, teammates, domain knowledge, workflows.
Varun Vummadi
>> Yeah.
John Furrier
>> You're starting to see more of an end to end workflow environment.
Varun Vummadi
>> Yeah. We're already doing that actually. Whenever a customer calls us, we take actions on behalf of the customers like canceling your credit card, canceling your airline, all those things. We are already doing that for some of our customers. As the models keep getting better and better, our customer support agents or any agents that we build in per se keep getting more smarter and smarter.
John Furrier
>> What's a good tell sign for someone who's interfacing with a chatbot or agent. What's a good agent look like? And what's a bad agent look like?
Varun Vummadi
>> Of course. The good agent, I think the fundamental principles are it should get you, the resolution of time should be significantly lower. If a human is taking 10 minutes to resolve it, it should resolve in generally less than three minutes because it has more context of you. It has also context of your past conversations. The second one big thing, this a lot of companies misses. If people ask for a human, just hold them for one thing that you can chat with AI, then you can just transfer them. You need to educate people to use AI. You cannot say that block, I'm not going to transfer you. That's going to get customers more frustrated. What we are seeing is a more of a generalistic approach. When people say agent, agent, agent, you can just say that, "Hey, I can help you resolve a billing issue if you have it." And the people might still want or prefer to a human, it should just transfer it.
John Furrier
>> And so you want to engage them with a little bit of a light touch?
Varun Vummadi
>> Yes.
John Furrier
>> And then see if they can progress further.
Varun Vummadi
>> Yes. Our fundamental belief is AI customer support is better than humans, and people should only use it if they believe it's better, and people should always have a way to human-
John Furrier
>> So your mission is to prove it.
Varun Vummadi
>> Yeah.
John Furrier
>> Give them some use cases.
Varun Vummadi
>> Yeah.
John Furrier
>> Okay. Connect you in one second, first question.
Varun Vummadi
>> Yeah.
John Furrier
>> Yeah. And then next thing you know, they're in a little non-
Varun Vummadi
>> I don't want them to be in a doom loop where they're like, "Oh my God, I hate Giga's AI. Yeah, it works, that it never transfers me to human."
John Furrier
>> So you guys got a 60 plus million Series A, big number, although I just saw Naveen Rao got a $400 million seed round, but whatever they call it, it's still probably priced rounds. You got to have customers, obviously have some success. Share some of the customers. What's the profile? What are the sizes? Can you give us a little taste of what the deployments look like, customers?
Varun Vummadi
>> Of course. DoorDash is a public name. And we are also working with some of the biggest financial services companies in the world and biggest telcos. Five big financial service and five big telcos. We want to specifically get closer to it. And we're also seeing a lot of traction mid-market and SMBs as well. But primary big ROIs have been being driven at massive public financial services companies and telcos. As you know, the upper expense of support and compliance are insane where even like one-
John Furrier
>> Scope the size of the OpEx, just order of magnitude.
Varun Vummadi
>> Yeah. For one company that we are working with, the support spend is 1.8 billion a year.
John Furrier
>> Wow. I mean, one little percentage point.
Varun Vummadi
>> Yeah, it's a lot of money.
John Furrier
>> A lot of money. All right. So talk about the team. How big is the company? What are you guys doing? What's the status?
Varun Vummadi
>> Yeah. We are 25 people right now trying to hit 100 or 150 by end of next year. We've been growing so fast. We're adding so many people recently. I was just joking with my team, I'm having hard time remembering everyone's names because every week there are so many.
John Furrier
>> A new person. All hands every day.
Varun Vummadi
>> Yeah.
John Furrier
>> That's great growth. I mean, it's fun to watch the multiplier, and this is what I like about this market right now, is you can have five people, 10 people. The multiplier effect has come up a lot on my CUBE interviews where the old 10X engineer used to be the thing we talked about, but now you have the 10X business person. The lines of businesses are deploying these solutions. When you go back on the cloud generation was DevOps was like the SREs. They're the ones running all the infrastructure, IT and then that replaced with cloud. Now you've got cloud scaled up. Now you've got AI scaling, so the scaling laws on AI are significant. Share your views on the scaling benefits, what you start seeing when you scale up. What are some of the scale laws you're following?
Varun Vummadi
>> Yeah. One thing is with AI, their models are already knowledgeable enough. They're solving math or MPI problems. Don't you think they can solve support? I think there is a significant gap in adoption of intelligence to adoption. I think that is what we're trying to solve. But I do think that there is some truth that as computer scaling, the intelligence is not scaling in parallel. It has slowed down compared to like the earlier times. But on a high level, there is a very big gap in adoption. I mean, a Math Olympiad, you have a kid who can do Math Olympiad problems, you should be able to do support much more than those things.
John Furrier
>> And again, what you're playing at is you're playing at the scale with the infrastructure, which is great, but also now the data. One of the things I liked about what came out of Amazon Web Services, AWS re:Invent this year was they introduced their half-baked Nova model and said, "Bring your own model to the table, get a frontier model without paying the price." Because in a lot of these use cases, you don't need the large language, you don't need the internet to solve a customer support problem where the domain is the customer's business.
Varun Vummadi
>> Yes.
John Furrier
>> That's not even in the models. So the roles of models are becoming, there's more digitalization, there's more customization. There's almost a frontier wave coming on custom models.
Varun Vummadi
>> You're 100% correct on this. The reason is you don't need to pay that much. Fundamentally, everything boils down to price. You don't need to pay a frontier model price to solve support. We are already starting to provide this thing to our customers, basically an auto fine-tuning capability. We take all the data that we generate from an OpenAI model or everything. We can just take it to try and a smaller model, which directly reduces-
John Furrier
>> Faster. Less GPUs. You can use XPUs, you use compute. You don't need to have... We saw that with DeepSeek. I mean, they got around H100s, 200s, because they had only H100s.
Varun Vummadi
>> Yeah.
John Furrier
>> So what they did is they just made it efficient. So I think efficiency is coming fast.
Varun Vummadi
>> Yes.
John Furrier
>> What's your advice for your customers and prospects now for this year? 2026 is the year of getting that value, practical. I was talking to Ali Ghodsi from Databricks here who's in the same chair you're in. And I was like, talking about AGI and he's like, "What's your AGI?" I had to ask him the AGI question. And he says, "Well, if you asked me in 2018 what AGI was and you showed me OpenAI today, I'd say that's AGI." He's making a point which is that it's pretty damn good right now. And then he said, "We're just working on solving practical stuff right now." And it's very pedestrian to say that, but we're in an era now where the hype is great. There's no real bubble in my mind. It's certainly bubblicious, but there's real value, and so this is where the money is.
Varun Vummadi
>> Yes.
John Furrier
>> Talk about that practical approach and what's the action items for your customers, prospects? What should it be doing? What should they be doing this year?
Varun Vummadi
>> Yeah. These are the obvious action items. I think any one of us would suggest: coding, support, and any ops adjustment use cases. They can just go through the P&L statement and take a look at ops and where there are like a lot of humans. If there are a lot of humans essentially they have returned policies, they can probably automate and make it significantly efficient using AI.
John Furrier
>> So target operations.
Varun Vummadi
>> Yes.
John Furrier
>> Look at the OpEx budget. Where's the spend, labor involved, tackle that first.
Varun Vummadi
>> Yeah, because-
John Furrier
>> Where's the pain point with customers? Where's the churn?
Varun Vummadi
>> Yep.
John Furrier
>> So sales, marketing, coding.
Varun Vummadi
>> Yeah.
John Furrier
>> Ops.
Varun Vummadi
>> Yes. Ops. The biggest option really is support. So support is obvious knockdown. There is one interesting thing I'm seeing with financial services customers is at the end of the call, if we know the customer is happy, we can upsell them a credit card or loan or anything. And you thought it would be crazy thinking that who would even buy from AI? If AI says that, "Hey, dude."
John Furrier
>> They're pissed off. I support, waiting on loan.
Varun Vummadi
>> But people are buying. I'm seeing real numbers there that for an exchange firm, for a trading firm, what I've seen as a crypto trading firm, I've seen an inactive customer getting active and them depositing more than $100,000 in the account backend, so AI is able to sell.
John Furrier
>> They got a dope hit. It's like a dopamine hit. I got my support taken care of. Great, I'm happy.
"While I got you on the phone."
Varun Vummadi
>> Okay.
John Furrier
>> Got my attention. That's a revenue upside.
Varun Vummadi
>> There's a revenue upside significantly. It's not with every single customer though. It's only with happy... You cannot oversell as well. It's a very like a thin line of.
John Furrier
>> Statistical conversion.
Varun Vummadi
>> Yeah, statistical conversion.
John Furrier
>> You say, "Okay, based on the numbers, we may have 15% affinity towards a nurtured progression," or just say thank you. Yes. Five stars, five stars are five stars, or get an NPS score, get that validation, but also so you got to keep the customer happy and revenue potential. All right. Varun, what's next? Put a plugin for the company. What are you looking to do? Obviously you're growing superfast, you're hiring. What are you optimizing for? What's your goals?
Varun Vummadi
>> For next year, the big goal is to, you are also right on point on this, just to scale fast and sell some of these massive customers and also like expand to addition markets and start proving and more and more efficiency gains.
John Furrier
>> And bring all that AI to the biz, your ops.
Varun Vummadi
>> Yeah.
John Furrier
>> You already have it done, you're pretty lean.
Varun Vummadi
>> We are trying to do it very aggressively in-house as well. What can we cut down?
John Furrier
>> Varun, great to chat with you. I'm so happy you came on, expert here. Again, we have our own mixture of experts here in theCUBE. Bringing all the action to you here from the New York Stock Exchange, of course, with our Palo Altos is theCUBE, 16 years covering tech. I've never seen it this good in terms of the hype cycle matching the demand curve. Unlike the internet, which is evolutionary, this is a transformative time and action items are identify the ops, look at the OpEx budgets, where's the labor, pick a use case, knock down some wins. This seems to be the theme. Varun, thanks and congratulations to your company and good luck. We'll keep in touch. I'm John Furrier, Host of theCUBE here at the NYSE, part of our NYSE Wired program. Thanks for watching.
>> Welcome back everyone, theCUBE. I'm John Furrier, Host. We here in the NYSE Studio of theCUBE. Of course, we have our Palo Alto studio connecting Silicon Valley and Wall Street. Tech and money are coming together. It's part of our mixture of expert series. We have an expert here who's doing amazing gen AI work in voice and generative content. Varun's here, he's the CEO, Co-founder of Giga. Varun, thanks for coming into theCUBE and to our NYSE studio on the East Coast. Part of our NYSE Wired program. Thanks for coming on.
Varun Vummadi
>> Thanks a lot for having me.
John Furrier
>> We were talking before we came on camera about the cool agents that are coming out, obviously avatars, real life, humans, voice, the role of voice. I mean, it's pretty well documented. All you got to do is get on X, get on Reddit, get on the internet. Voice is the killer app for AI. We all use it on text. We use it with OpenAI. We use it for all the tools. But voice is actually a killer app inside enterprises so you're starting to see money making platforms coming in, leveraging the data. So very hot area you're in right now, so congratulations. Explain for the folks what you guys do, then we'll get into some of the deeper dive.
Varun Vummadi
>> Yeah, sure. For a little bit about Giga, we have recently ran a $61 million Series A. And what we do is we build a customer support agents for enterprises like DoorDash, et cetera, using us. And the biggest adoption, as we are seeing, as you were mentioning is on voice, because you eliminate hold times. None of us like being on hold so we are essentially killing it.
John Furrier
>> I saw a stat from the DoorDash guy, I saw a quote you guys have on the website. I know they're a customer. He said that their drivers could do a variety of different things. And one of the comments he said was, "I want to build a solution that gets them an answer as fast as possible."
Varun Vummadi
>> Yes.
John Furrier
>> Huge important value for him because their company, they have independent drivers and delivery. You don't want to be on hold. Someone pays, it didn't go through. I mean, all kinds of shit happens. They got to speed that cycle down.
Varun Vummadi
>> It's not just drivers, none of us want to be on hold as well. There's a stat, on average, a human spends 48 days of their lifetime on hold. It's like you spend 48 days of your life on hold.
John Furrier
>> We've all been there. We're not saying you do your email. "Oh, okay. Where am I?" You're connected an hour later.
Varun Vummadi
>> One of my inspirations was while earlier in move to US, I called IRS. They had me on hold for six hours. Six hours. I don't know how many times you have dialed IRS.
John Furrier
>> And then of course, it's like fifth hour in it can disconnect.
Varun Vummadi
>> Yeah, disconnect. Oh my God. And you need to keep jumping things. They need to keep transferring you, yeah. It's a pretty big motivation for me to give people... You can just let people get what they want in a very shorter amount of time.
John Furrier
>> Yeah. I mean, during COVID, one of the things that jumped out at me with AWS was Connect, which was their service. It was very clear that low hanging fruit was the data that the corpus of data from all the support. But now that generative AI is here, almost all the successful beachhead positions start with use cases that a company could knock down.
Varun Vummadi
>> Yeah.
John Furrier
>> Call center, customer support. They have domain data. They're getting new data so you have all the elements of good AI right there. Talk about where they're at right now. What are some of the things look like for you guys? What are some of the use cases? What's some of the value? Can you share some stories?
Varun Vummadi
>> Yeah, of course. The biggest, you are correct on point. The biggest enterprise adoption is happening on support and coding because they're like the first use cases that people are able to knock down. But I'm seeing a clear picture of companies who want to automate more things, especially financial services. The biggest op expense of these companies, if you look at, are support and compliance. They have a lot of people taking a look at these things. It's essentially like policies. And instead of a human, AI takes a look at the policies and automates the task. And then again, HR comes next, but support and compliance, companies spend billions of dollars a year on this.
John Furrier
>> They can make or break a customer relationship on this, and this is where I think the frustration turns into opportunity because I'm a consumer. I'll go to a better support whether it's an airline or whatever.
Varun Vummadi
>> Yeah.
John Furrier
>> If I can get an answer quicker, I'm happy. All right, so take me through what's going on inside the enterprise, because obviously there's a couple things that we can review that go back last year and a year and a half, two years. RAG has been popular. Search, that's search.
Varun Vummadi
>> Yep.
John Furrier
>> Same discovery. What's the answer? And then customer support. And both are grounded on the fact that these enterprises have data.
Varun Vummadi
>> Yes.
John Furrier
>> And it's domain specific to their world.
Varun Vummadi
>> Yes.
John Furrier
>> And there's databases everywhere, so you got the combination of these siloed databases. You now have a point of support with data. Existing data, knowledge systems, but also new data. So voice is actually fresh data too, which is a training opportunity, a reinforced learning opportunity. What's your take on all that?
Varun Vummadi
>> Yeah, you're on point. The way when we start working with an enterprise, the way we do it is we consume all the existing data, which sometimes is not even properly documented or written. We consume all the existing data and build this support policies. We keep iteratively improving on it by learning from their existing customer support agents. That is how we do reinforcement learning to keep reading improvement those policies. It's not just where we land. Let's say we ran at 60% resolution rate or 70% CSAT. We want to show companies a clear pathway to reach from 98% or 99%. The only way we can do it is iteratively learning from the humans on what we don't have and what they have.
John Furrier
>> Okay. So talk about the competition. Obviously, this is a highly competitive environment.
Varun Vummadi
>> Yes.
John Furrier
>> Everyone sees this as the low hanging fruit.
Varun Vummadi
>> Yeah.
John Furrier
>> Good use case. What are you guys doing that's different? Can you share the secret sauce?
Varun Vummadi
>> Of course. The big differentiator is a lot of our competitors take Palantir-like approach. It's became very common in AI, which is a forward deployed engineer. We give you engineers. We sit with you in a company and we do it. While it's a great approach, it takes a lot of time to see results. It roughly takes four to six months for them to deploy. We generally go live in less than two weeks. This is, I'm talking with some of the massive enterprises in the world.
John Furrier
>> Two weeks.
Varun Vummadi
>> Two weeks. And we are a product company. We don't have already deployed engineers. As you were mentioning, we take all the data and build the policies and we tell you the agent, you can chat with it and we change it and it can go live.
John Furrier
>> You're basically doing some heavy lifting on behalf of the customer.
Varun Vummadi
>> Yes. We're basically automating the forward deployed engineer path. That is why we can go faster. That's one big thing. The other big thing is iterative improvements. No matter where you land, let's say if you land at 40% resolution rate, because you don't have all the data to resolve the issues at the start, it will iteratively get you to 98% as fast as possible. We also have this exciting launch coming up in a week where we help customers without APIs also to get the benefits faster.
John Furrier
>> All right, so now the agents are coming.
Varun Vummadi
>> Yeah.
John Furrier
>> Where are agents going to fit into your plan? How do you see the agentic piece? Because now that you have some of these capabilities, it's more than a chatbot. There's past completion, teammates, domain knowledge, workflows.
Varun Vummadi
>> Yeah.
John Furrier
>> You're starting to see more of an end to end workflow environment.
Varun Vummadi
>> Yeah. We're already doing that actually. Whenever a customer calls us, we take actions on behalf of the customers like canceling your credit card, canceling your airline, all those things. We are already doing that for some of our customers. As the models keep getting better and better, our customer support agents or any agents that we build in per se keep getting more smarter and smarter.
John Furrier
>> What's a good tell sign for someone who's interfacing with a chatbot or agent. What's a good agent look like? And what's a bad agent look like?
Varun Vummadi
>> Of course. The good agent, I think the fundamental principles are it should get you, the resolution of time should be significantly lower. If a human is taking 10 minutes to resolve it, it should resolve in generally less than three minutes because it has more context of you. It has also context of your past conversations. The second one big thing, this a lot of companies misses. If people ask for a human, just hold them for one thing that you can chat with AI, then you can just transfer them. You need to educate people to use AI. You cannot say that block, I'm not going to transfer you. That's going to get customers more frustrated. What we are seeing is a more of a generalistic approach. When people say agent, agent, agent, you can just say that, "Hey, I can help you resolve a billing issue if you have it." And the people might still want or prefer to a human, it should just transfer it.
John Furrier
>> And so you want to engage them with a little bit of a light touch?
Varun Vummadi
>> Yes.
John Furrier
>> And then see if they can progress further.
Varun Vummadi
>> Yes. Our fundamental belief is AI customer support is better than humans, and people should only use it if they believe it's better, and people should always have a way to human-
John Furrier
>> So your mission is to prove it.
Varun Vummadi
>> Yeah.
John Furrier
>> Give them some use cases.
Varun Vummadi
>> Yeah.
John Furrier
>> Okay. Connect you in one second, first question.
Varun Vummadi
>> Yeah.
John Furrier
>> Yeah. And then next thing you know, they're in a little non-
Varun Vummadi
>> I don't want them to be in a doom loop where they're like, "Oh my God, I hate Giga's AI. Yeah, it works, that it never transfers me to human."
John Furrier
>> So you guys got a 60 plus million Series A, big number, although I just saw Naveen Rao got a $400 million seed round, but whatever they call it, it's still probably priced rounds. You got to have customers, obviously have some success. Share some of the customers. What's the profile? What are the sizes? Can you give us a little taste of what the deployments look like, customers?
Varun Vummadi
>> Of course. DoorDash is a public name. And we are also working with some of the biggest financial services companies in the world and biggest telcos. Five big financial service and five big telcos. We want to specifically get closer to it. And we're also seeing a lot of traction mid-market and SMBs as well. But primary big ROIs have been being driven at massive public financial services companies and telcos. As you know, the upper expense of support and compliance are insane where even like one-
John Furrier
>> Scope the size of the OpEx, just order of magnitude.
Varun Vummadi
>> Yeah. For one company that we are working with, the support spend is 1.8 billion a year.
John Furrier
>> Wow. I mean, one little percentage point.
Varun Vummadi
>> Yeah, it's a lot of money.
John Furrier
>> A lot of money. All right. So talk about the team. How big is the company? What are you guys doing? What's the status?
Varun Vummadi
>> Yeah. We are 25 people right now trying to hit 100 or 150 by end of next year. We've been growing so fast. We're adding so many people recently. I was just joking with my team, I'm having hard time remembering everyone's names because every week there are so many.
John Furrier
>> A new person. All hands every day.
Varun Vummadi
>> Yeah.
John Furrier
>> That's great growth. I mean, it's fun to watch the multiplier, and this is what I like about this market right now, is you can have five people, 10 people. The multiplier effect has come up a lot on my CUBE interviews where the old 10X engineer used to be the thing we talked about, but now you have the 10X business person. The lines of businesses are deploying these solutions. When you go back on the cloud generation was DevOps was like the SREs. They're the ones running all the infrastructure, IT and then that replaced with cloud. Now you've got cloud scaled up. Now you've got AI scaling, so the scaling laws on AI are significant. Share your views on the scaling benefits, what you start seeing when you scale up. What are some of the scale laws you're following?
Varun Vummadi
>> Yeah. One thing is with AI, their models are already knowledgeable enough. They're solving math or MPI problems. Don't you think they can solve support? I think there is a significant gap in adoption of intelligence to adoption. I think that is what we're trying to solve. But I do think that there is some truth that as computer scaling, the intelligence is not scaling in parallel. It has slowed down compared to like the earlier times. But on a high level, there is a very big gap in adoption. I mean, a Math Olympiad, you have a kid who can do Math Olympiad problems, you should be able to do support much more than those things.
John Furrier
>> And again, what you're playing at is you're playing at the scale with the infrastructure, which is great, but also now the data. One of the things I liked about what came out of Amazon Web Services, AWS re:Invent this year was they introduced their half-baked Nova model and said, "Bring your own model to the table, get a frontier model without paying the price." Because in a lot of these use cases, you don't need the large language, you don't need the internet to solve a customer support problem where the domain is the customer's business.
Varun Vummadi
>> Yes.
John Furrier
>> That's not even in the models. So the roles of models are becoming, there's more digitalization, there's more customization. There's almost a frontier wave coming on custom models.
Varun Vummadi
>> You're 100% correct on this. The reason is you don't need to pay that much. Fundamentally, everything boils down to price. You don't need to pay a frontier model price to solve support. We are already starting to provide this thing to our customers, basically an auto fine-tuning capability. We take all the data that we generate from an OpenAI model or everything. We can just take it to try and a smaller model, which directly reduces-
John Furrier
>> Faster. Less GPUs. You can use XPUs, you use compute. You don't need to have... We saw that with DeepSeek. I mean, they got around H100s, 200s, because they had only H100s.
Varun Vummadi
>> Yeah.
John Furrier
>> So what they did is they just made it efficient. So I think efficiency is coming fast.
Varun Vummadi
>> Yes.
John Furrier
>> What's your advice for your customers and prospects now for this year? 2026 is the year of getting that value, practical. I was talking to Ali Ghodsi from Databricks here who's in the same chair you're in. And I was like, talking about AGI and he's like, "What's your AGI?" I had to ask him the AGI question. And he says, "Well, if you asked me in 2018 what AGI was and you showed me OpenAI today, I'd say that's AGI." He's making a point which is that it's pretty damn good right now. And then he said, "We're just working on solving practical stuff right now." And it's very pedestrian to say that, but we're in an era now where the hype is great. There's no real bubble in my mind. It's certainly bubblicious, but there's real value, and so this is where the money is.
Varun Vummadi
>> Yes.
John Furrier
>> Talk about that practical approach and what's the action items for your customers, prospects? What should it be doing? What should they be doing this year?
Varun Vummadi
>> Yeah. These are the obvious action items. I think any one of us would suggest: coding, support, and any ops adjustment use cases. They can just go through the P&L statement and take a look at ops and where there are like a lot of humans. If there are a lot of humans essentially they have returned policies, they can probably automate and make it significantly efficient using AI.
John Furrier
>> So target operations.
Varun Vummadi
>> Yes.
John Furrier
>> Look at the OpEx budget. Where's the spend, labor involved, tackle that first.
Varun Vummadi
>> Yeah, because-
John Furrier
>> Where's the pain point with customers? Where's the churn?
Varun Vummadi
>> Yep.
John Furrier
>> So sales, marketing, coding.
Varun Vummadi
>> Yeah.
John Furrier
>> Ops.
Varun Vummadi
>> Yes. Ops. The biggest option really is support. So support is obvious knockdown. There is one interesting thing I'm seeing with financial services customers is at the end of the call, if we know the customer is happy, we can upsell them a credit card or loan or anything. And you thought it would be crazy thinking that who would even buy from AI? If AI says that, "Hey, dude."
John Furrier
>> They're pissed off. I support, waiting on loan.
Varun Vummadi
>> But people are buying. I'm seeing real numbers there that for an exchange firm, for a trading firm, what I've seen as a crypto trading firm, I've seen an inactive customer getting active and them depositing more than $100,000 in the account backend, so AI is able to sell.
John Furrier
>> They got a dope hit. It's like a dopamine hit. I got my support taken care of. Great, I'm happy.
"While I got you on the phone."
Varun Vummadi
>> Okay.
John Furrier
>> Got my attention. That's a revenue upside.
Varun Vummadi
>> There's a revenue upside significantly. It's not with every single customer though. It's only with happy... You cannot oversell as well. It's a very like a thin line of.
John Furrier
>> Statistical conversion.
Varun Vummadi
>> Yeah, statistical conversion.
John Furrier
>> You say, "Okay, based on the numbers, we may have 15% affinity towards a nurtured progression," or just say thank you. Yes. Five stars, five stars are five stars, or get an NPS score, get that validation, but also so you got to keep the customer happy and revenue potential. All right. Varun, what's next? Put a plugin for the company. What are you looking to do? Obviously you're growing superfast, you're hiring. What are you optimizing for? What's your goals?
Varun Vummadi
>> For next year, the big goal is to, you are also right on point on this, just to scale fast and sell some of these massive customers and also like expand to addition markets and start proving and more and more efficiency gains.
John Furrier
>> And bring all that AI to the biz, your ops.
Varun Vummadi
>> Yeah.
John Furrier
>> You already have it done, you're pretty lean.
Varun Vummadi
>> We are trying to do it very aggressively in-house as well. What can we cut down?
John Furrier
>> Varun, great to chat with you. I'm so happy you came on, expert here. Again, we have our own mixture of experts here in theCUBE. Bringing all the action to you here from the New York Stock Exchange, of course, with our Palo Altos is theCUBE, 16 years covering tech. I've never seen it this good in terms of the hype cycle matching the demand curve. Unlike the internet, which is evolutionary, this is a transformative time and action items are identify the ops, look at the OpEx budgets, where's the labor, pick a use case, knock down some wins. This seems to be the theme. Varun, thanks and congratulations to your company and good luck. We'll keep in touch. I'm John Furrier, Host of theCUBE here at the NYSE, part of our NYSE Wired program. Thanks for watching.
John Furrier
Varun Vummadi