In The Arena by TechArena - Data Insights: Enterprise AI with Adity Dokania

Episode Date: June 25, 2026

In this episode of Data Insights, Allyson Klein and Jeniece Wnorowski sit down with Adity Dokania to explore how enterprise AI is evolving from experimentation into production. The conversation focuses on the operational challenges that emerge at scale, including data trust, governance, infrastructure constraints, and cross-functional ownership.

Welcome to Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome in the arena. My name is Allison Klein, and today is a Data Insights episode, which means we're here with Janice Norowski. Welcome, Janice. It's always great to see you. Hi, Allison. Thank you so much. It's great to be back. Janice, I know that every single episode we do is delightful, but this one is particular interesting to me. Tell me what the topic is today and who you've brought with you as our guest. I'm very excited as well. Today we have Adity Dapnya from Ken Show. And we're going to be talking

a lot about all things AI as we have been. And let's just say it right in. Welcome, Adity. Can you share a little bit about Ken Show and your role at the company? Yeah, thanks for having me, Allison. Thanks, Jimis. So Ken Show is an AI in Data Analytics company. And And what makes it a fascinating place to work is that we were acquired by SNP Global, which is one of the world's most established financial intelligence companies. So you've got this taught up DNA with the speed, the experimentation, the willingness to try new things, operating within a very large, very regulated enterprise. And I am the director of cloud and infrastructure and security,

and a big part of my role is bringing frontier AI capabilities to life in ways in which they actually work for our enterprise clients. So Adity, you sit at a unique intersection at Kencho working with frontier capabilities for usable large enterprises. How do you describe the scope of what you're responsible for today and where's most of your attention go? So I often describe my job as being a translator, but in both directions. On one hand, I'm working with teams that are pushing the boundaries of what AI can do. And on the other hand, I'm sitting with enterprise compliance and security teams who have very real constraints. They have legacy systems.

They have regulatory scrutiny, risk aversion, and that's justified. So a big part of what I do is figure out how. to bridge that gap, not by watering down the AI, but by designing the right guardrails and integration pathways so that something genuinely powerful can be deployed and it is responsible. So in terms of where my attention lies, honestly, it's on the operationalization of these AI capabilities. Getting something to work like in a demo is one thing, but getting it to run reliably at scale, inside existing workflows with audit trails, with governance, all of that baked in has been a large part of where my focus lies.

For when you're seeing across enterprise clients, what feels most different about AI adoption right now compared to even a year ago, especially as organizations think about scale and durability? The biggest shift in my mind that I'd point out is, that the conversation has moved from, is it even possible to being accountable, right? Like a year ago, a lot of the enterprise leaders were in exploration mode, running pilots, figuring out what the technology could do, getting comfortable with the vocabulary even. But now, I think boards are asking for ROIs.

Businesses are asking for something within production. And actually, this is a healthy shift, right? but it creates real pressure because the infrastructure and the governance aspects of it are not there yet. What I'm seeing is this tension between leadership that wants to move fast and the technical and risk teams that are basically trying to lay the groundwork of how to correctly deploy all of this into the ecosystem. Many enterprises today really want to move quickly with AI. I mean, everyone wants to see their return. investment, but they are constrained by existing data infrastructure. In your opinion, where do you see

the biggest kind of friction points as AI becomes more central and core two systems? Yeah, data quality is definitely one I keep coming back to. And I don't mean clean data. I mean data that's trusted, that has lineage, that people inside the organization actually agree upon because there is just so much data in any enterprise or organization. And then there is the infrastructure there that you pointed out. A lot of enterprises are running core systems that were not built for AI, right? So we are retrofitting AI into these environments and even those companies are trying to figure that out. But beyond data and infrastructure, I also say that the organizational structure is something that causes friction.

AI doesn't fit neatly into like an existing team boundaries. It requires like collaboration between like legal, compliance, business, engineering, all of them to come together simultaneously. And that's where I feel like the speed at which something can get to market gets inhibited. Now, Kentra was often early to experiment, but enterprise clients expect production grade outcomes. How do you decide which AI capabilities are ready to move forward, and which need more trying before they can responsibly be deployed?

Yeah, this is a good one. I mean, there is a framework I followed even before AI has tweaked it a little bit for the purposes of AI. But the three dimensions that I look at is observability. One big piece is like, can the system be monitored once we put it out there in the world? The second one is reproducibility or explainability. We subtracted it to explainability because we know that these systems are non-deterministic, so we won't be able to ask and get back the same answers,

but can we explain how it arrived to an answer? Why did the system make a decision that it made? And the third one is, can we build resilience into the system? What happens when the system is wrong, right? Because it will be wrong sometimes. The question is whether it fails in a way that's recoverable and detectable. So those are the three things I keep coming back to. Got it.

You know, as AI workloads grow, how are enterprises rethinking where computation and data kind of live? How are they deciding between cloud services and bringing back certain workloads on-prem? Yeah, this is, I would say, genuinely an evolving conversation. And what I find interesting is that the answer is becoming less binary. A few years ago, the narrative was, pretty clearly everything moves to the cloud period.

It's the new way of doing things, and that's where we're going to live. Now we're seeing a more nuanced picture, sensitive data, specially regulated industries, prompting real questions about where inference actually happens. Right. So you're seeing a kind of gravitational pull back towards a hybrid model where like maybe you could do training in your cloud environment, but the inference and where the decision-making happens, our most sensitive data stays closer to home. So that's kind of what I'm seeing with our clients today.

But again, like I said, it's so evolving and so new that we still have to talk, figure it out, do the right thing. From your experience, how does working with enterprise data change the way teams approach model training in iteration compared to more controlled or research-oriented environments? Yeah, it's very different actually. in a research environment, you have a certain luxury of your data set is defined, your success metrics are clear, and iteration is relatively clean, right? Enterprise data is much messier in every dimension.

It's incomplete, it's inconsistent across systems, and it has historical biases baked into it that reflect past business decisions rather than the ground truth. So I feel like people who understand the data, the domain experts of this data are never in the same room where these decisions are being made about the data. So one of the biggest shifts I see with enterprise AI development is investing more upfront in understanding the data and partnering with these SMEs or domain experts to validate what the model is actually learning on its data. So when AI initiatives struggle

inside large organizations, what are some of the common gaps between what the technology teams filled and what enterprises are actually prepared to support operationally? Yeah, I think it comes down to clear communication and having fragmented responsibilities. The gap in my mind is around ownership, right? Who owns the system after its lives? Who monitors it? Who trains it when the model drifts? AI cannot be built by one team and maintained by another team.

It is a continuously evolving process. And it's changing so fast that engineering teams need to work together and communicate with one another. And that in my mind I feel like is missing sometimes. Sometimes teams think that it is my responsibility to build, but then once it's out in the world, some other team is going to maintain it and figure out all of the longer term considerations around it. As we look forward to the rest of the year, what signals are you watching for that will indicate whether enterprise AI adoption

is moving from pilots and series of isolated experiments to really being embedded into the way businesses operate? The signal I care about most is whether AI is showing up in operating metrics rather than in project reports. When a CFO starts talking about AI-driven efficiency in earnings cause, not as a future initiative, but as a current contributor that's embedding AI into its system.

The other signal I watch is talent when enterprises start hiring for AI operations and AI governance roles, not just data scientists and engineers, right? We've been seeing the rise of data scientists and engineers who work on AI for a while, But as I said in my previous answers, the shift in focus is around governance and around operational AI. And when I start seeing those numbers stick in terms of hiring, those are the signals I'm currently watching for. So to wrap up, what are some of the most important things that your partners are focusing on today to kind of health enterprises be ready for the next phase of AI adoption? I think few things. First, as I've mentioned repeatedly, governance infrastructure, helping enterprises build the staff holding around AI, like the model gateway,

the observability layer, the audit trails, basically the boring stuff that makes everything as sustainable. The second is use case prioritization. So working with leadership to identify where AI can have disproportionate impact in sequencing initiatives in a way that builds internal confidence. And the third is meeting the clients where they are. integrating into the existing workflows, working closely with our clients and building partnerships, listening to their needs instead of just handing them a chat interface and asking them to figure it out.

I feel like has all of our partners thinking about. Adani has been a real pleasure talking to you today. I am sure that our listeners are eager to learn more about what you're doing at Ken Show and how they can engage. He has shared where they can find information about the company and your solutions and how to engage your tool? Yeah, a great question to end on. Obviously, any information on the company,

any of the releases that we do on our products are on the press release at sbglobal.com forward slash press releases. We also have Kencho's blog on Kencho.com slash blogs. So any of the new inventions, anything new that we have been working on and gets posted there as well. And you can always reach out to me at Aditi at Kensha. Thank you so much for being here today.

And Janice, that wraps another episode of Data Insights on the Arena. It's always a pleasure. Thanks for being here. Thank you, Alison. Thank you, Alice. Thank you, thank you. Thanks for joining Tech Arena. Subscribe and engage at our website, Techorina.coma.

All content is copyright by TechRena.