Pure Accelerate 2026 | Ashish Gupta, Everpure

Clips
More from Pure Accelerate 2026

Ashish Gupta

VP & GM, Data Management

Everpure

play_circle_outline Drivers for building 1touch: GDPR, misplaced data, AI requiring context

play_circle_outline Making Data AI‑Ready: Unified Discovery, Classification, and Governance Across On‑Prem, Cloud, Mainframes with Everpure

play_circle_outline Handling unstructured data growth and risks of sharing via spreadsheets or messaging

play_circle_outline Data-Centric Governance for Real-Time AI: Metadata, Sovereignty, and Compliance to Prevent LLM Data Leaks

Info
Transcript

Ashish Gupta, Everpure

Ashish Gupta

VP & GM, Data Management Everpure

In this interview from Pure Accelerate 2026, Ashish Gupta, general manager of data management at Everpure, joins theCUBE's Christophe Bertrand and Alison Kosik to discuss the path to AI-ready data through discovery, context and enterprise-wide governance. Gupta traces the origins of 1touch.io — the company he founded before joining Everpure — to three converging pressures: the rise of GDPR, the security risk of misplaced data and the context requirements of modern AI workloads. He outlines the foundational steps enterprises must take to achieve AI-ready data:... Read more

explore Keep Exploring

What motivated the founding of 1touch? add

What does it mean for data to be "AI-ready," and how should organizations discover, govern, and prepare data — including mainframe and unstructured sources — to ensure compliance when using it for AI? add

How does your platform discover and classify data across mainframe, on‑premises, and cloud environments, and how does it add contextual information (such as business-process relationships and timelines) to that data—particularly unstructured data—to ensure AI uses it safely? add

Where should a chief data officer or chief technology officer begin when building toward true data intelligence, and what does the roadmap or initial steps look like? add

bolt Powered by CUBE AI

Ashish Gupta, Everpure

search

Alison Kosik

>> Welcome back to Pure Accelerate 2026. We are live in Las Vegas. I'm Alison Kosik joined alongside by Christophe Bertrand, and we're about to bring in a guest who is going to be talking about a topic that you know really well about, context, semantics, and the path to data intelligence. Right in your wheelhouse, Christophe.

Christophe Bertrand

>> Exactly. An area that I have researched over the years and certainly covered also when I was on the product side of things.

Alison Kosik

>> Absolutely. So let's dig into it with Ashish Gupta. He's the general manager of data management with Everpure. Welcome to theCUBE. So great to have you with us.

Ashish Gupta

>> Thank you, Alison and Christophe. It's great to be here on theCUBE with you.

Alison Kosik

>> Ashish, you led 1touch before joining Everpure, and you've spent years at the intersection of data intelligence and enterprise AI. Talk about why you built 1touch, and what is it that you were looking to solve?

Ashish Gupta

>> Yup. We started off with 1touch with three ideas in mind that were driving our decision to build the company. The first one was GDPR was taking over, and a big part of GDPR was really understanding where your data was and then managing it appropriately. The second one was I used to run a company that did crowdsourced cybersecurity, and 80% of our, if you will, bugs that were submitted for our customers were based on data that was singing where it should not have been singing. In other words, was misplaced. And then the third one was we are seeing this revolution with AI, and AI consumes data from everywhere, but it requires context. To do the right thing, you need to understand your data so that it can build the context, so that AI can be more accurate. So those are the three drivers for us.

Alison Kosik

>> Yeah.

Christophe Bertrand

>> Right. And I think, actually, there's even more than that, because you think about it. You mentioned GDPR, and, of course, there's always going to be a new one, as well as some AI-related compliance regulations that are popping up all over the place.

Ashish Gupta

>> Yeah.

Christophe Bertrand

>> What's interesting here is that, yes, of course, AI needs data that has the right context so it can feed the model correctly, but I think there's also a very strong compliance component, because-

Ashish Gupta

>> Totally....

Christophe Bertrand

>> you can't just use any data. So what is the right data to use for AI? What are the characteristics?

Ashish Gupta

>> So you're absolutely right, Christophe, and it shows that you've got that background in this. So the first part of this is understanding your data, and the three things you need to do to understand your data is first know where it is. Just most importantly, see where your total visibility is of the data so that you can determine what's the right data to feed to the AI models. And third, do it in a compliant way. It's governed by certain rules. Some rules are internal in nature, but then every day, you're seeing new rules pop up from a compliance sovereignty point of view and an AI sovereignty point of view.

Christophe Bertrand

>> Right. Exactly. And this is the difference between maybe compliance and its cousin governance. Governance is a set of rules that, as an organization-

Ashish Gupta

>> Exactly right....

Christophe Bertrand

>> you will add, very likely encompassing a number of compliance regulations as well.

Ashish Gupta

>> That's right.

Christophe Bertrand

>> All right. So everything changes with AI. Now, we have agents making decisions, happily going, transacting with individuals or with other agents as a matter of fact. How does it change the game?

Ashish Gupta

>> It changes the game first by really understanding what the business process is and then applying certain specific intelligence around that business process to make it happen a lot faster. This gives us a tremendous amount of opportunity to do other things and releasing a lot of resources to enable more innovation to happen. So I feel AI is going to be very, very productive for everyone in the market, but it's going to be very productive only if you've got the right data context driving that accuracy.

Alison Kosik

>> Why is it so hard for enterprises to achieve AI-ready data? I mean, enterprises have more data than ever.

Ashish Gupta

>> Yeah.

Alison Kosik

>> Why is it so hard to achieve the AI-ready data, and why is it we are seeing these AI projects stall?

Ashish Gupta

>> Yeah. It's a great question. So I think it's a multipart answer to that. First one is, when you think about automation, there used to be an old book that people used to talk about. I think Tim Davenport said it. Don't automate, obliterate. So the first thing is that you need to understand what you want to let AI do. From that second perspective comes around is, what is the data that is going to make it more accurate in doing what it needs to do? Third is it needs to be accepted culturally within the organization. And then fourth, you need to continue to make sure that it's accurate and learning as it moves along. So when you take all of these things together, in addition to the fact that costs can burgeon if you don't do it correctly, both from an opportunity cost of view and the true cost of running AI, it becomes quite a large task if it's not done in a coordinated manner.

Christophe Bertrand

>> Right. And what's interesting here is there a couple of angles to this, but obviously, you want to be covering all of the data. Otherwise, it doesn't work, right? Which means that you have data in SaaS environments. You have data on multiple storage systems, Everpure and others. You have data that maybe you don't even know about, but you have-

Ashish Gupta

>> Exactly right....

Christophe Bertrand

>> which is probably more problematic, which is a compliance exposure.

Ashish Gupta

>> Yup.

Christophe Bertrand

>> So would you say that, in a sense, with the Everpure Data Intelligence, not only are you getting contextualized data for AI at scale across the whole set of data that your enterprise runs, but you're almost getting compliance by default, because you're doing all the work.

Ashish Gupta

>> Exactly. That's a fantastic observation. When you think about AI-ready data, people think about it as, first, you need to understand the data. Second, you need to make sure it's compliant and governed in the right way, and third is you feed the right data from a context point of view. If you take these three LEGO blocks, a new LEGO block can be put on top called AI-ready data. But to really understand your data, you need to discover it. As you said, Christophe, it sits anywhere. It could be on-prem, in the cloud, in mainframes. Interestingly, 68% of all transactional data in financial institutions still goes through mainframes.

Christophe Bertrand

>> Not a surprise.

Ashish Gupta

>> It is pretty incredible, and we are the only company that actually goes out and discovers and classifies the data on mainframes just like we do on on-prem and cloud. Second is, there's a lot of data that is unstructured and structured. I would submit unstructured data is growing uncontrollably and much faster than structured data. You could just take, Alison, a Salesforce report, download it into a spreadsheet, and send it to three friends in the company, and all of a sudden, you've got very, very sensitive information that is unstructured in nature being shared across the organization. But the AI tool gets a hand on that, it might not be used in the best way possible. Right? Once you've gotten discovery and classification done, that's when the magic of Everpure Data Intelligence starts. We then take that discovery and classification. We start to add context, context based on business processes that use the data, context based on how that data is related to other data. So it's an invoice to a certain person paid at a certain time. That relationship and timeline is super important for AI to know when it's going to be processing some sorts of information related to that invoice.

Christophe Bertrand

>> So let's talk about that a little more, because you've just given a very simple example. In reality, there will be multiple agents potentially running, leveraging the data at the same time for transactional purposes. And obviously, you don't want to wait too long. You want that to happen real time. You want it to happen on the actual dataset that is in production. How do you reconcile that with doing all that work of context of classification, of maybe raising the flag to say, "Well, hang on, hang on. You cannot use that data here, because it's protected by GDPR or it's protected by HIPAA," what have you?

Ashish Gupta

>> A very, very fond analogy comes to mind. One of 1touch's board members, Bernard Gavgani, who used to be the CIO of BNP Paribas, said to me, he goes, "It's not about just AI. It's about the rocks you have to jump as you cross the river." And I remember this in a small little Airbnb in Paris that he was sharing this with us as we were determining, "How do we build this knowledge graph, this knowledge graph of context?" And I give him a lot of credit, because it helped us drive our road map, to your point, Christophe. The first thing is you need to be able to take data at rest just as much as data in motion and ensure that you've put the right tags and the right information from a metadata point of view on that data.

Christophe Bertrand

>> Right.

Ashish Gupta

>> So when the time comes, you can make the right decision on it. Second is, you need to know which business processes use what kind of information. And as Charlie said in one of our keynotes here, is that you need to have a source of truth. The source of truth is not your dad's old source of truth, which used to be an application, a walled garden of information that said all customer data just falls in a CRM system. The source of truth today is going to be how you extract that data out, become data-primary and data-centric, and make that data available through metadata tags at the right time, including the location of the data, including the compliance requirements to that data, and specifically the relevance of that data for those AI tools.

Alison Kosik

>> If a CDO or a CTO is going to go back to their organization and wants to start building toward true data intelligence, where do they start, and what does the path look like from there?

Ashish Gupta

>> They start with Everpure Data Intelligence, of course. So my take on this is, they should really, at first, make sure they understand where all that data is. So Everpure Data Intelligence goes out in the first week to two weeks, give you a really good sense for what you could call your asset surface of data. It's, again, regardless of what storage device it's sitting on, on-prem or cloud, and definitely structured or unstructured, data in motion, data at rest. So you get that flexibility. Within the first month, we see our customers finding information that they thought was protected but is actually available. So one of the main use cases most of our customers use us for in day one is to find production data sitting in non-production environments. Now, while you might think there's a compliance requirement for that, there is. There's a much more dangerous, lurking fear, is that non-production data, sometimes mass, sometimes not mass, sometimes protected, sometimes not protected, gets into a LLM.

Christophe Bertrand

>> Right.

Ashish Gupta

>> That is very dangerous. It's scary, because, first, it should not go there. Secondly, it might not have the right context, because the data was changed by a production engineer who was trying different things out. So we see a lot of our customers, our CTOs and CDOs, go in and say, "First step, remove all the non-production environments of all the production data. Second, you classify the data." Interestingly, only 15% of our organization's data is considered sensitive at that time. Does not mean that the historic data is not important, but that's what's sensitive. So identifying that and making sure that's available is super important. I'll just give you an example. One of our chief data officers, a major hotel company, said, "I can build a loyalty program. No problems." It's a few relational databases. Christophe, you and I wrote a few of those in our time. Right? And you can say, "Here are the three tiers of my loyalty program." But now, someone is making a decision using an AI agent to make a reservation. Happens to be in the United States going to the United Kingdom. All of a sudden, there's data leaking out, which is saying, "This is Christophe living in the United States looking for a reservation in the UK with my credit card information and address," toxic combination of information. But there's a law called ARGOS, which says U.S. resident information should never be held in a non-U.S. hotel. To that end, we look at that information and say, "You cannot be sharing this," making it easier for the AI tool to use the metadata about Christophe, in this example, to make the reservation without releasing all the other sensitive information. So now, you become part of the solution while making sure that the reservation is done, while making sure the revenue is attained, and ensuring you didn't have a compliance violation.

Christophe Bertrand

>> So let's follow up on that, because you've mentioned two job titles, which I would say, a couple of years ago, three years ago maybe, may have been foreign to Everpure.

Ashish Gupta

>> May have been?

Christophe Bertrand

>> Foreign to Everpure. Chief data officer. CTO, probably not so much. CTO, of course, but chief data officer is an interesting one. I think the personas, it's a discussion we're having with Chad as well on the product management side of things. The decision-making is changing. There is already this changing of the guard in terms of cyber resiliency versus CIOs, CISOs. Who's really in charge? Are the teams converging? And they are.

Ashish Gupta

>> Yup.

Christophe Bertrand

>> But I see a third component, which is the data component also converging now in the context of AI, because you have to have those three legs of the stool for it to work. How does that change how you approach the market?

Ashish Gupta

>> Yup. It's a fantastic question, and three legs of the stool is a fantastic way to put it. So my take on this is that large enterprises are already going down what we call alliance decision, because data has become so important that they need to make sure the CIO, CISO, and chief data officer are very much in line with the decision they're making on what solution they're going to build. We see this in every one of our customers. We've got over 50 Fortune 100 companies across the United States and Europe, and it's interesting to me that you'll always have a use case that starts the decision. At a major credit card company, the use case came with GDPR, and the person was actually asking us to say, "We get 7 to 800 DSAR requests, data subject access requests," which basically means, "Tell me where all my end user information is, and tell me that it was actually deleted." It used to take this company 21 person-weeks to address that.

Christophe Bertrand

>> Wow.

Ashish Gupta

>> They had over 100 petabytes of information. I bet in the last two years I've been working with them, that's grown to 120 petabytes, because data is growing so fast. 21 weeks were taken down to 30 seconds with our platform.

Christophe Bertrand

>> Wow.

Alison Kosik

>> That's amazing.

Ashish Gupta

>> That was the start. How did we do it? We gave total visibility. All of a sudden, the CISO comes in and says, "Wait a second. I need to use this exact technology to say, 'Am I able to use my data leak prevention tools and enable it to get the intelligence of what's sensitive and not sensitive and have one policy across the entire organization managed by Everpure Data Intelligence and ensure that no leak prevention happens?'" which then opened the door for the chief data officer to come in and say, "Well, wait a second. I could use this information and the context about that end user and why that end user is important to this credit card company and enable better AI results to come out of it." So we see this alliance come together. The second part of it is that, for Everpure, we've built some very, very deep relationships in almost 15,000 customers, and I'm really proud of my new brethren and sisters that I'm going to be building the business up with Everpure Data Intelligence, because they've built these relationships in multiple different parts of the organization. And because we enjoy, I think, a 84 net promoter score, we see our inside personas already proposing new personas to come work with us on a daily basis. So we feel the three legs of the stool are going to come together with Everpure being the one place that they can sit on.

Alison Kosik

>> This has been a great conversation. Thanks so much for stopping by theCUBE. Loved listening to your expertise today, Ashish.

Ashish Gupta

>> Thank you so much for the opportunity to speak with you. Good meeting you again-

Alison Kosik

>> Thank you....

Ashish Gupta

>> and look forward to future conversations too.

Alison Kosik

>> Thank you. And you've been watching theCUBE, the leader in live technology coverage and expert analysis on technology. We'll be right back.