Ataccama Puts Data Trust at the Center of Enterprise AI

Hey guys, thanks for the tour. We're here with Martin Zahumenský, who's the newly appointed CEO for Ataccama, and we're talking about, well, digital trust, data trust, AI trust, all the kinds of trust that are out there. But Martin, welcome to the show.

Hello. Thanks, Mike. Thanks for having me.

You are not a newcomer to Ataccama. I think you were the CTO before this, but now you're taking over as CEO. But let me ask you something up top and just from the very beginning, but it seems like one of the things that's going on in AI is it's shining a light on this whole issue of data.

Do we trust it? And what's the validity of the source? Because, well, the AI is basically only as good as the data that was first shown it.

So what's your sense of what's going on here, and do we have a new appreciation for data management quality and maybe there's a renaissance going on? I think so. We see agents just amplifying the issues which were there for the last 20 years.

Isn't too much change there. Just like as the AI agents get implemented and they start talking to each other, the whole problem with data is starting to be more visible. We were used to people looking on the data.

They have the judgment. They have a history with the company. They understand the whole context.

Agents, a lot of the times, are making autonomous decisions, so the impact of potentially making an incorrect decision because the data fed to the agents or the data generated by the engines are incorrect are just magnifying all the issues and potentially impacting the company. So I think we're seeing some renaissance there, and we're seeing many more companies now investing in the data foundations because that's probably the biggest blocker point in implementation of AI agents on the enterprise level. So where exactly does your company fit in that equation?

What is it that you guys are doing and kind of, if you wouldn't mind taking a minute, just kind of figuring out a little bit, where does Ataccama sit in the general landscape of data management? We position ourselves as the trust layer for your data, which essentially means that to-- the problems you see with the trust in the data, it's coming from many different angles, and it always was. People looking on the reports, they feeling something is not correct.

It might be tons of different stuff happening on the background. So the pipelines maybe haven't run. I don't know, certain region was not including in the report or in the latest version of the report.

It might be the reference data used to calculate the reports or feeding the application data, it's incorrect. It might be data quality, just that the elements are not captured or the data are not captured at the input properly. " And we believe that to solve this problem, you really need several different pieces in place.

So we basically cover everything from data cataloging, understanding where the data sits, what the data means, how to use the data, over to data quality, so making sure the critical data elements in the companies are covered with the expectations and that those data actually are getting improved over the time, to observing the state of the pipelines over to reference data management, master data management. So the whole point is to be able to solve this problem, and especially in the AI world, you need to be able to cover this whole landscape to give holistic information about which data can be trusted and which data should be used by the agents, which should not be used by the agents, and generating the answers to the questions people have. How does that work?

Because you're touching on what's a sore point for a lot of people where theoretically, we want to make data-driven decisions, but a lot of the business leaders know where the data came from in the first place, so they're suspicious, and they rely more and more on their instincts rather than what the data might be saying. So how do I validate that the data is actually relevant and correct? It's not too much change in this area.

I think just having proper data governance in place, which is describing the sources, the meaning of the data in those source systems, how they are used at the end, for what reports or what applications they are feeding. I think all this stuff is needed in the AI world as well, probably even more. We also see that there is some theory, and a lot of the vendors is pushing that the agents will be creating on one space, somewhere in Snowflake or Databricks on other technologies.

But we also see that due to the fact that the big organizations we work for, especially in the financial institutions or insurance, they have tons of systems. They've been building the data landscape over years. So there is a lot of fragmentations in those data, and you have to be able to describe the whole ecosystem and be able to shift the data quality and this whole stuff towards live, towards the source systems.

And a lot of the agents are actually being built towards the source system. So that's I think where Ataccama really differs from the other technological vendors and where we are adding value. Of course, humans have been reading those reports for a while, but as we go forward, AI agents will be reading those reports and acting on that.

So is that part of this new trust virtuous cycle that we need to create because those AI agents will be acting on that data at machine speed, and they may not stop to think twice. Yeah. I think that's the biggest change we see in general.

I think we'll have the biggest impact on the whole data management area. We were moving a lot into the MCP direction, being able to expose everything Ataccama does to the agent so they can make the decisions as autonomously as possible and get all the metadata about which data can be trusted, which might not be fit for that specific use case. So I think this whole composability of the AI ecosystem and being able to plug into that, it's super important because a lot of the stuff And we see that a lot of people try to automate everything within their platform, which we also do as well to just be able to generate the data quality rules at scale and these things.

But then when we want to solve the bigger issues and automate the processes in the organization, I think it's super important that the AI can actually compose the different tools, including Ataccama, to be able to solve that problem. So nice example is regulatory compliance. I think AI is pretty good at understanding generally what regulatory compliance means, and then if it would be just acting within the tool, like Ataccama, you only can go so far.

But if you can provide that whole information about lineage, data quality of the assets, what assets we are having in the organization, then AI can solve much more complex systems. So I think that's where we are focusing a lot now to be able to work in the AI automation within the enterprises and how to put actually data management on the level where it needs to be. So under your direction, what will be different about the company?

What's your priority list? I wouldn't say a lot of things are super different. I think it's a lot about listening to customers and being able to understand where the whole AI is moving and where are the kind of the white spots where new solutions are, or where you need to fit as a solution in that broader AI ecosystem play.

So definitely a lot of focus how we can bring trusted data to the AI agents and how we can make AI agent which can be actually trusted more within the organizations. So that's where the product engineering effort will be going. Then obviously reducing time to value, especially in the whole AI ecosystem, it's super important.

So focusing a lot on how we can be faster in delivering value to our customers and just strengthening the whole strategy we had about the unified platform because we strongly believe that to make this AI ecosystem better, you need to have these pieces I was just talking about a minute ago accessible and point solutions only can go so far. You need also the breadth to be able to solve that whole problem about AI or the data being ready for the AI. " I think in the overall AI hype, I think a lot of companies is just focusing on localized problems, like there is a lot of people experimenting with AI locally and trying to solve this specific problem, create this application.

I think where the real value is coming of, it's redesigning the processes, rethinking how the whole organization works, what roles you should have, how the processes should work, and not too many are focusing on the higher-level, outcome-driven kind of thinking around the AI. And that goes to data management as well. " But at the end, I think the biggest leverage comes when AI can touch all the data because obviously the power is much more immense than power of people or us as individuals.

So I think focusing on being able to have the data in the shape that AI can touch anything and then can make bigger decisions and automate broader problems is the thing they should be looking on instead of just looking on localized kind of issues which do not have lever at the output of the company at the end. So one of the issues that you hear people talk about is that there is compliance mandates and auditors and all those things, and if I'm making a decision and it goes wrong, they want to know what my data sources are, and it seems to me, though, that going forward, they too are going to have AI tools that are going to look through our workflows and our processes and call out where we violated something at machine speed. So is the whole nature of how we think about compliance and trusted data, is that conversation going to change?

I think from some perspective, yes. I think the typical compliance we see or the regulatory requirements we see, I think portion of that will stay probably as it is. Then in the AI, there are whole new problems created as you just stated.

So there will be new requirements on being able to audit log and prove how the AI was deciding what data were used for those decisions. So I think from portion, definitely it will be changing. So ultimately, what's your best advice to folks out there that are managing data and they're probably looking at AI with both a sense of excitement and fear?

The excitement comes from the fact they will probably get more value out of the data than ever, and the fear is that the data will be accessed more aggressively than ever. So how do I approach this, or what should I be thinking about? I think I would say always think about the broader picture.

As I said, a lot of the companies is driving data modernization, just being able to have an infrastructure where you can easily implement the agents and they can get access to the data in the company and provision the answers through a question or through reports, but through conversational AI and analytics on demand. I think what I would not underestimate is the preparation work needed to be able to execute on this. So I think the AI is still in the stage where a lot of people is experimenting, the value is being judged.

I think the preparation work needed for AI to be successful is a real thing you should be focusing right now on because it takes time. A lot of these processes, cleaning the data, understanding the data, documenting the data properly takes time. So I would put definitely an effort during those data modernization plays on this part of the problem as well, not just focusing on getting new shiny data management platform out there without understanding if the data can be used and it's fit for the purposes or for the use cases you're trying to solve through the AI agents.

Ultimately, am I going to need some type of AI or AI agent to clean up the data to be consumed by other AI agents? That's true. A lot of focus we are putting in the platform is obviously on AI agents, which can automate this problem and can better discover the issues in the data, propose solutions, how you can solve that problems or cleanse them automatically.

So that's definitely the place where the world is going and the whole data management is going. Rule SDQ can probably get on steroids with AI compared to we were used to do before. So that's where big focus of the product team is going into.

All right. Folks, you heard it here. Hey, no matter what era it is, it's all about the data one way or another.

Hey, Martin, thanks for being on the show. Thank you, Mike. All right.

Back to you guys in the studio.