KiE Square Proves AI Fails Without Strong Data Engineering

There's a quiet truth in the AI industry that most companies learn the hard way: the model is rarely the problem. The data feeding it is. Dr. Kamaljit Anand, Managing Partner and Chief Data Scientist at KiE Square, has spent close to two decades watching brilliant data science models stumble — not because the math was wrong, but because the pipes carrying the data were broken. His message is blunt and refreshingly unfashionable in an era obsessed with the next big model: if your data engineering is weak, your AI is going to fail. Poor data in, poor decisions out. No algorithm can save you from that.

KiE Square is a data analytics and AI consulting firm built on a "techno-functional" philosophy — the idea that you need to understand both the business problem and the technical machinery to actually solve anything. Dr. Anand co-founded the company alongside batchmates and juniors from IIM Ahmedabad, where he completed an integrated MBA and a PhD in quantitative marketing. That dual training, business plus deep technical depth, ended up shaping the entire way the firm approaches its work.

From Predicting Cricket Matches to Pricing Engines

Dr. Anand's path into analytics started long before it was a buzzword. His curiosity began with something playful — trying to predict the winners of sports series using probability, combinations, and joint and associative reasoning. That itch to predict things, to find the signal in the noise, eventually carried him to IIM Ahmedabad and into the very early days of the global data analytics lifecycle.

His career moved through Gallup, where psychology met customer engagement and quantitative market research, and then Fractal Analytics in its formative years, where he did notable work in the FMCG and consumer goods space along with financial services. When he set up KiE Square, the work was first statistics-driven, then evolved into data science and analytics, and today is firmly AI and machine learning driven. It's been a long arc, and along the way he watched the field's center of gravity shift in a way that most people still underestimate.

The Gap Everyone Missed: The Data Backbone

This is the heart of Dr. Anand's argument, and it's worth sitting with. In the early days, solving a problem was a long, drawn-out affair. You had time to conceive a solution, design it, and implement it carefully. But as the field matured, KiE Square noticed something that kept tripping up even talented teams. Data scientists were burning enormous amounts of time wrestling with broken pipelines instead of doing actual science. The plumbing was eating the value.

So the firm made a deliberate bet. Instead of just hiring data scientists, they started cultivating people who understand data science deeply but operate with a data engineering mindset. That shift, he argues, is what lets you build strong data lakes, robust data warehouses, and genuinely useful functional data marts — the foundation that everything else rests on. They paired this with strong data visualization, because a data science output only matters if a human can see it, feel it, and turn it into a day-to-day decision.

A $120 Billion Pricing Problem, Solved Daily

If you want to understand why data engineering matters at scale, consider one of KiE Square's clients — a Fortune 25 pharmaceutical distribution corporation in the US. This company runs price engines that support more than $120 billion worth of business. Every single day, those engines have to generate a unique price for every product going to every customer.

Think about the combinatorial explosion for a moment. If you have 10,000 SKUs and 10,000 retailers, the number of price combinations is staggering, and a fresh price has to be generated for each one every day. The price might be identical to yesterday's, but an algorithm still has to approve or disapprove it. That's a data science problem on the surface, but it's impossible without a serious data engineering pipeline underneath — one that pulls data from the source, cleans and shapes it, maps historical pricing, benchmarks against competitor data, and then pushes approved prices out to the decision-making systems where a hospital, pharmacy, or doctor's clinic can actually see them.

Part of the work was migrating old SAS-based systems into a modern PySpark design where Python becomes the engine. For some price engines, migration wasn't feasible, so they continued running on SAS. The lesson Dr. Anand draws is that everything has to flow together — the pipeline, the data science, and the visualization — or the whole thing breaks. A legacy standalone tool has to become a true enterprise tool so decisions happen at the enterprise level, not in isolated pockets.

KiEVerse: Productizing Decades of Experience

After hundreds of implementations, KiE Square did what experienced services firms eventually do — they productized their hardest-won expertise into software. That product is KiEVerse, a digital marketing intelligence suite that turns the firm's marketing analytics experience into a near plug-and-play, software-as-a-service offering. And with AI baked in, it learns and adapts on its own, which makes execution and on-the-fly changes far easier for clients.

The reason KiEVerse exists points to a very modern headache. A large consumer goods company used to sell through two or three marketplaces. Today it has to push product across quick commerce, e-commerce, physical retail chains, and direct-to-consumer channels all at once — and drive traffic to each one through top, middle, and bottom-of-funnel spending. Figuring out how much you spent versus how much you sold across all those fragmented platforms, in their native formats, is a nightmare. You need it unified in a single design layer, in near real time, so you can see spend and sales together and compute ROI on the fly.

Because this is a specialized job, KiE Square treats it as a dedicated marketing data lake — or marketing data mart — delivered "service as a software." That framing matters. It's not entirely plug and play, Dr. Anand is honest about that, but it's close enough to make life dramatically easier for clients who would otherwise drown in platform fragmentation.

When Milliseconds Decide Everything

Some of the most demanding pipelines KiE Square builds operate at speeds that leave no room for sloppiness. Take an e-commerce recommendation system. You want to recommend a product based on a customer's purchase history — but only if the inventory actually exists. Recommend something you can't physically deliver, and you've broken the customer's trust along with the back-end supply chain.

To get this right, you need a real-time streaming pipeline — something like Apache Flink — aggregating user click streams as they happen. The same product might be selling across twenty cities simultaneously, with inventory continuously depleting. You have to build a clean user vector for consumption and bring it down to a hundred-millisecond or sub-hundred-millisecond latency before deciding whether a product can even be recommended.

The stakes climb even higher in financial markets. In a stock exchange, countless trades happen at any instant for a single instrument, with algorithms entering and exiting positions just like human traders. The data stream is brutally fast. To handle it, KiE Square reaches for tools like Kafka for streaming, and dbt or Apache Airflow for orchestration. As Dr. Anand puts it, "the pipelines are very serious pipelines" — because if you don't make them tight, the failures carry real consequences.

AI: Disruptor and Enabler at Once

So where does the current AI wave fit into all this? Dr. Anand refuses to give a one-sided answer. AI, he says, is both a disruptor and an enabler. It disrupts certain things companies used to do, and it enables entirely new services built on top of it. But he's clear-eyed about where it falls short — and where the industry has stumbled.

The mistake many teams make is treating AI as a point-to-point solution. You bolt an AI product onto a narrow problem without giving the system a real design, and you end up with something tactical that never becomes strategic. For multi-billion-dollar enterprises, that's a non-starter. They don't want to gamble on point solutions; they want robust enterprise systems. And because they're still figuring out which AI use cases truly matter, their decision cycles remain long — which has slowed the industry even as it has opened up a huge field of opportunity.

One striking data point from the conversation: today, only about 20% of an organization's relevant data sits inside the organization. Roughly 70–80% lives outside it. The challenge — and the opportunity — is bringing that external data in and building a system robust enough to handle it. That's why enterprises increasingly want AI data lakes that integrate everything, not another isolated tool.

A Smarter Way to Price a Data Product

Pricing a service is relatively easy — you can anchor it to man-hours. Pricing a data product is harder, and Dr. Anand's approach is elegant precisely because it ties the price to the value being moved. For KiEVerse, the model links directly to the scale of the client's business.

The Future: Infrastructure, GCCs, and Sovereign AI

Looking five to ten years out, Dr. Anand believes infrastructure becomes more important than ever. Services will still matter — data engineering especially will stay strong and critical — but visualization will cede a lot of ground to AI that can generate it faster, even if accuracy still lags. He also points to the rise of Global Capability Centres and the large-scale deployment of GPUs, where the sustainable use of energy will be a defining concern.

His sharpest conviction, though, is about sovereign AI. He argues it's going to be extremely critical, because it determines how an organization's data flows — how much stays within their own design versus how much sits inside an external large language model that may not have access to their proprietary frameworks. His prescription is concrete: develop or deploy small language models (SLMs) specific to your industry, host them locally, build MCP designs and your own AI frameworks around them, and keep proprietary data in a protected, isolated, slightly siloed environment.

For KiE Square itself, the road ahead has two lanes. First, keep strengthening the services around data engineering and reach more clients and industries that need the right foundations for their AI ambitions. Second, keep spinning out product companies in the mold of KiEVerse. Dr. Anand hinted at two more in the pipeline — one in the government space, focused on sophisticated algorithmic designs for fraud identification and tax systems plagued by pilferage and evasion, and another aimed at the pharma sector. Each new product is meant to augment the mother company, the services business that is KiE Square.

A Necessary Reality Check on AI

What makes Dr. Anand's perspective valuable is that it lands at exactly the moment the industry is sobering up. A year ago, everyone was shouting "AI this, AI that." Now there's a reckoning — privacy concerns around models trained on proprietary data without licensing, safety scares like AI agents deleting production databases, and a growing recognition that the technology isn't as infallible as the hype suggested. Dr. Anand isn't dismissive of AI; he's emphatic that it's required. But he insists the missing piece is the right data engineering framework. Get the data foundation right, and AI delivers. Get it wrong, and no model in the world will save you.

Key Takeaways

• The data backbone matters more than the model: Clean pipelines, data lakes, and data marts are the foundation. Poor data quality in means poor outputs out — no algorithm can compensate.
• Hire for a data engineering mindset: KiE Square's edge came from cultivating people who understand data science but build with engineering discipline, freeing scientists from broken pipelines.
• Scale is brutal and unforgiving: From a $120B daily pricing operation to sub-100ms recommendation latency, real enterprise AI lives or dies on pipeline rigor.
• KiEVerse productizes hard-won expertise: SpendVerse, BrandVerse, and MarketVerse turn fragmented marketing data into unified, near real-time decisions delivered as "service as a software."
• Price the product to the value it moves: Tying fees to spend or stream volume keeps unit economics simple and aligned with the client's business.
• Sovereign AI is the next frontier: Industry-specific SLMs hosted locally, with proprietary data kept isolated, give organizations control over how their data flows.

About the Guest

Dr. Kamaljit Anand is the Managing Partner and Chief Data Scientist and a co-founder of KiE Square (KiE Square Analytics), and the founder of KiEVerse.ai. He holds an integrated MBA and a PhD in quantitative marketing from IIM Ahmedabad, and brings nearly two decades of experience delivering analytics solutions to Fortune 500 corporations across financial services, retail, consumer goods, and telecom. His work spans marketing decisions, customer management, risk management, and pricing optimization. Before founding KiE Square, he served as a practice director for client delivery at Fractal Analytics and earlier worked at Gallup.

KiE Square is a data analytics and AI consulting firm built on a techno-functional philosophy that pairs deep business understanding with strong data engineering and data science capability. The firm builds enterprise-grade data lakes, warehouses, and pipelines, and delivers data science and visualization solutions for global clients. Its digital marketing intelligence product suite, KiEVerse, comprises SpendVerse (spend and ROAS optimization), BrandVerse (martech audits and digital brand-equity scoring via Digi-Cadence), and MarketVerse (cross-platform e-commerce market intelligence via Meta-360), with additional product companies in the government and pharma sectors in the pipeline.