Every retailer now has an AI story. Personalized homepages, chat assistants that answer at midnight, demand forecasts that adjust by the hour, store associates handed a model’s recommendation instead of a hunch. Most of these stories are still pilots. The demo lands in the boardroom, the budget gets approved, and then the project stalls somewhere between the proof of concept and the checkout page, defeated by data that does not line up, systems that will not talk to each other, and governance questions nobody wanted to answer first. The distance between what retail AI promises and what it reliably ships is rarely a modeling problem. It is an architecture problem, and it lives in the parts of the business that never make the keynote.
Shrinivas Jagtap has spent more than 20 years in exactly those unglamorous parts. A senior technical architect and a Senior Member of the IEEE, he has built the data platforms, integration layers, and enterprise systems that sit beneath the retail and supply-chain software many companies run their operations on, and more recently the generative AI features, chat assistants, intelligent search, document summarization, and predictive insights, layered on top of them. His book, The Retail Intelligence Stack: A Practical Playbook for Transforming Customer Experience, Operations, and Growth, is a case for paying attention to the layers beneath the intelligence, because that is where most retail AI quietly succeeds or fails.
That background shapes the book. It does not read like a trend report on generative AI in retail, and it does not treat the technology as a finished product you simply buy and switch on. The author has clearly watched promising pilots collapse under messy data and has had to build the foundation that keeps the next one standing, and that experience runs through every chapter. Its 239 pages work through the full stack: the data layer, enterprise architecture, commerce and customer experience, store operations, marketing, and the governance and security that decide whether any of it can be trusted at scale.
Where the intelligence actually lands
What keeps that idea from staying abstract is how concretely the book applies it across the retail floor. Personalization, commerce, marketing, store operations, and fulfillment each get treated as a place where the same discipline pays off: clean, shared data underneath, well-designed interfaces between systems, and a clear view of which decisions a model should make on its own and which still need a human. Jagtap is careful not to sell a single silver bullet. He walks through where generative AI genuinely helps, a shopper asking a store’s catalog a question in plain language, a merchandiser summarizing a quarter of reviews in seconds, and where it mostly adds risk if the plumbing is not ready for it.
For all the architecture talk, the book keeps its eye on the customer. The reason to fix the data layer is not tidiness. It is that a shopper feels the difference between a retailer that knows them and one that only claims to. Jagtap makes the case that customer experience is the visible output of everything underneath: a recommendation that actually fits, a service answer that is right the first time, an inventory promise the store can keep. When the stack is sound, those moments feel effortless. When it is not, the customer is the one who absorbs the mistakes, and no amount of front-end polish hides a broken foundation for long. That framing, experience as the surface of good engineering, is what keeps the book from reading like an IT manual and gives it something to say to the merchants and operators who will never write a line of code.
Lessons from the field
The book’s ideas are not borrowed. They come straight out of work Jagtap has led. Building a multi-tenant data platform for a large distributor, he designed the ingestion, validation, and sharing layers that turned scattered records into a single trusted source, cutting cloud costs by roughly 40% and support-engineering effort by about 75% while pushing data accuracy up sharply. On a separate integration program for an industrial equipment rental company, one unified data model replaced a tangle of point-to-point connections, cut the volume of system calls by more than half, and raised data accuracy to around 95%. Neither project makes a good demo. Both are exactly the foundation the book insists retail AI cannot skip.
Higher up the stack, his more recent work is the generative AI that customers and staff actually touch: assistants that answer natural-language questions against enterprise data, search that reads intent rather than keywords, summarization that turns long documents into something a busy manager can act on. Earlier in his career he built a responsive B2B commerce platform that let a marketing-services firm spin up custom online stores and run the full order lifecycle, from request through fulfillment, which helped that client grow revenue by about 25%. The pattern across all of it holds. The interface gets the attention. The architecture underneath is what makes it dependable.
The road ahead
The timing suits the argument. Retailers are under pressure to show a return on their AI spending, regulators are starting to ask how automated decisions get made, and the easy pilots have mostly been done. What is left is the harder, structural work of making intelligence dependable at scale, which is precisely the work the book maps. Jagtap does not promise that good architecture makes AI easy. His claim is more useful than that: the retailers who win with AI will be the ones who treat it as a stack to be engineered rather than a feature to be bought, and who do the quiet work at the bottom before chasing the impressive work at the top. For anyone trying to move retail AI out of the pilot phase, the book reads like a working blueprint for where to begin.
Leave a Reply