Beyond Chatbots: Dview bets on AI that can reason

Dview positions itself as a data and AI-first enterprise intelligence platform, focused on bringing data, knowledge, computation, and AI reasoning together. The company has grown roughly 4x over the last three quarters, while its technology has evolved from conversational analytics towards unified enterprise knowledge and Agentic AI. In this conversation with ET Digital, Kauts Shukla, Founder & CEO of Dview, explains why he believes intelligence is becoming a foundational layer of the enterprise stack, why enterprises can embrace this shift without rebuilding their existing infrastructure, and why the next generation of AI will be defined not by how much data organisations can process, but by how intelligently they can reason over it and turn that intelligence into action. Edited excerpts.

The Economic Times (ET): What was the problem or gap in the market that led you and Supratik Shankar to start Dview? What did you see in the way businesses were managing and using data that convinced you there was a need for a new kind of data platform?

Kauts Shukla (KS): Dview started with a fairly simple observation: enterprises had become exceptionally good at generating and storing data, but remarkably poor at turning that data into intelligence at the speed the business needed. I had spent years working inside data-intensive businesses and kept seeing the same pattern. A leadership team could have petabytes of data, sophisticated warehouses, multiple analytics tools, and large data teams, and yet a seemingly simple business question could still take days to answer.


The problem wasn't a lack of data or even a lack of computing power. The missing layer was intelligence - a system that could understand what the business was asking, identify the relevant data, reason across it and determine what actually needed to be computed. That became the foundation for Dview.

We started with the belief that the next generation of enterprise data infrastructure could not simply be about moving, storing, or visualising data. It had to understand the meaning behind the data and make that knowledge usable by both humans and AI systems.

ET: Dview describes itself as a Data-as-a-Service platform. What exactly does that mean for a business, and how does it simplify the journey from raw data to actionable insights?

KS: Data-as-a-Service, at least the way we mean it, is not another API sitting in front of a database. It's an attempt to change the relationship between a business and its own data.

Think about what a typical enterprise actually looks like. Transactional systems, a dozen SaaS applications, a few databases, a warehouse, a lake, and a set of flat files everyone has quietly agreed to ignore. Then a second layer for transformation, a third for analytics, and now a fourth for AI. Every one of those layers is defensible on its own. Put together, they create an enormous amount of infrastructure between a business question and a business answer.

Dview is trying to collapse that distance. We are building an intelligent data layer that ties together three things which have historically lived apart. The data, the computation, and the business knowledge. Someone should be able to ask what's happening in the business, why it's happening, and what they should look at next, without having to know how the plumbing works.

I want to be careful about one thing, though, because it's where a lot of the market is currently confused. Natural language is only the interface. It's the least interesting part of the system. The hard part is everything behind it. Understanding the intent, understanding the shape and semantics of the underlying data, working out what needs to be computed, executing that efficiently, and then returning something a CFO would be willing to put in a board pack.

ET: There are several companies offering data management, analytics and AI solutions. How is Dview different from the competition? What is your core technological or business advantage that would make a customer choose Dview over an established player?

KS: The distinction starts with where you believe intelligence belongs. A lot of enterprise AI today is still being built as an application layer sitting on top of an existing data architecture. Our thesis is different: intelligence needs to become part of the data infrastructure itself.

Dview is designed around the convergence of data, knowledge, computation, and AI reasoning. Instead of asking an AI system to simply retrieve information from an existing stack, we are building an architecture where it can understand enterprise semantics, reason across disparate sources, determine the most efficient way to compute an answer, and eventually orchestrate actions.

That matters because enterprise AI is not fundamentally a chatbot problem. It is a context, reasoning and trust problem.

An AI model may know a great deal about the world, but it doesn't automatically know what "active customer" means inside a particular bank, how that organisation defines profitability, which data it is permitted to access, or which business rules should govern a decision.

That enterprise context is where we see a major opportunity.

The next generation of AI systems will not simply retrieve enterprise knowledge. They will reason over it. That requires a unified knowledge architecture underneath the models, not another interface sitting above fragmented systems.

This is also why our work has increasingly moved towards Agentic AI systems that can investigate a question across multiple sources, reason through the evidence, generate insights and eventually take governed action.

The ambition is to make enterprise intelligence a core infrastructure capability rather than another application layered on top of the stack.

That thesis has also received strong external validation. Dview was selected as part of Google for Startups Accelerator: AI First, Class of 2025, one of 20 startups selected for the programme. We were subsequently selected for the Velocity – Kotak BizLabs Incubation Program at NSRCEL, IIM Bangalore, and more recently entered into a strategic investment and technology partnership with Grand View Research focused on combining unified knowledge architecture, Agentic AI and enterprise intelligence.

ET: Dview’s platform can centralise data from more than 100 sources and allows users to interact with data through natural-language queries. How does this work in practice, and what kind of business decisions can customers make faster because of this capability?

KS: We do connect to more than 100 sources, and I would say that part is table stakes. Every serious platform gets there, eventually. Typing a question in plain English is also easy to demo. I could build that in a weekend. The hard problem is making the answer correct, contextual, governed, and cheap to produce. All four, every time.

Take any large business. Customer records in one system, transactions in another, finance in a third, operational data spread across several more, and a warehouse that holds some but not all of it. A business head shouldn't need a mental map of that landscape. They should be able to ask why revenue fell in a particular region, which customer segment is quietly becoming unprofitable, what changed in unit economics last quarter, or which SKUs are dragging the margin down, and get an answer they can act on.

For the system to do that, it has to work out the intent, find the relevant data, understand how those tables relate, apply the company's own definitions rather than generic ones, generate the right computation and then check its own work. That final validation step matters more than people expect. An analytics system that is confidently wrong is worse than no system at all.

Which is why I resist the Text-to-SQL label. Text-to-SQL is a feature. Data reasoning is the opportunity. What we are working towards is a system that understands an enterprise's data environment roughly the way a very experienced data scientist who has been in the company for four years understands it.

The practical effect is on the economics of decision-making. Questions that used to need an analyst, a data engineer, and three rounds of back-and-forth become interactive. And here is the part I find most interesting. When the cost of asking a question falls close to zero, people stop rationing their curiosity. They start asking better questions. In my experience, that's where the real change in a company happens. Not in the tool, but in the surrounding behaviour.

ET: Who are your customers today? Which industries and types of businesses are adopting Dview, how many clients do you currently serve, and can you share a few examples of the kinds of data challenges you are solving for them?

KS: We work with enterprises, meaning organisations where data is already central to how the business runs, but is scattered across a complicated technology landscape that has usually been accumulated over a decade rather than designed.

The pattern we look for is specific. A lot of data, many systems, and a real distance between the people who own the data and the people who need answers from it. The industry varies more than the problem does.

A financial services firm may be trying to stitch together customer, transaction and risk information that has never sat in one place. A consumer or commerce business wants to see customer behaviour, revenue and operational performance in the same frame instead of three separate reports that disagree with each other. A large enterprise may have years of data across several warehouses, databases and applications and no unified intelligence layer over any of it. And then there's a category I find particularly interesting. Companies with a very modern stack who are simply spending far too much on compute, because the system is grinding through vastly more data than the question required.

None of these are BI problems in the old sense. They're enterprise intelligence problems.

ET: What has the business achieved in terms of revenue and growth? Can you share your current annual revenue/ARR, revenue growth over the last two or three years, and what your revenue target is for the next 12 to 24 months?

KS: The number I am happy to share, and the one I think actually says something, is that we have grown roughly 4x in the last three quarters. Not over a comfortable multi-year run, but in three quarters. That is the figure I would point anyone to, because it tells you the demand is present tense rather than historical.

Given that we only started operating commercially in December 2021, and that I was on my own for the first stretch of it, I am reasonably satisfied with where that puts us.

ET: You claim that Dview’s architecture can improve computing efficiency by 60% while offering 99.9% uptime and complete data control. What enables these efficiencies, and how do you ensure that the platform remains secure and reliable as customers scale their data operations?

KS: There's a reflex in this industry that says if it's slow, throw more compute at it. I think the opposite is becoming true. The smartest compute is the one you never have to run at all.

Most of our engineering effort goes into query intelligence, which means working out which data is relevant to a question at all, how it should be accessed, what can be pruned before we touch it, what actually needs to be calculated, and where that calculation should happen. None of this is glamorous work. It doesn't demo well. But at scale it's where the money is.

And that is not a small effect. A modest reduction in how much unnecessary data you scan compounds into a very large difference in the infrastructure bill once you're operating at enterprise volumes. That's why we spent our early years on the data and query architecture rather than rushing to ship a chat window on top of a warehouse. In real customer workloads we've seen this translate into up to 60% improvement in computing efficiency, while holding the existing SLAs. And those gains come from doing less, not from renting more.

ET: Both founders bring significant experience in building data and technology platforms, including work at OLA, OLA Electric, Vedantu, Moody Analytics and OakNorth. How has that experience shaped Dview, and where do you see the company in the next three to five years?

KS: The experience we brought into Dview gave us something particularly valuable: a view of enterprise technology from the inside including where conventional architectures begin to break down as data, complexity and scale increase.

Across mobility, financial services, edtech and other data-intensive environments, one pattern becomes very clear: the technology stack can become extremely sophisticated while the distance between data and decision-making continues to grow. That experience shaped our fundamental question.

What should enterprise infrastructure look like if AI is no longer just an application sitting on top of data, but an intelligence layer capable of reasoning across the enterprise? That is the direction we've been building towards.