This article was based on the interview with Salesforce’s Nitin Mangtani on how AI is evolving on-site search by Greg Kihlström, AI and MarTech keynote speaker for The Agile Brand with Greg Kihlström podcast. Listen to the original episode here:
For as long as we’ve had e-commerce, we’ve had the on-site search box. It sits there, a humble and often frustrating gateway to a brand’s entire catalog. We, as marketing and commerce leaders, have spent countless hours optimizing it, creating synonym lists, and building labyrinthine merchandising rules to guide customers toward a purchase. We’ve A/B tested its placement, its placeholder text, and the color of its button. Yet, for all that effort, it has remained a fundamentally blunt instrument. The experience is often a “fire and forget” transaction where the customer types a few keywords, gets a grid of products, and is left to their own devices. The dreaded “zero results found” page remains one of the most common—and most costly—points of failure in the digital customer journey.
But what if that transactional, often-broken utility could become the most intelligent, personalized, and conversational part of your brand experience? The shift we are witnessing is not an incremental improvement; it is a complete reimagining of digital interaction. Fueled by advancements in generative and agentic AI, we are moving from a world of rigid keyword matching to one of conversational discovery. In a recent conversation, I spoke with Nitin Mangtani, GM and EVP of Agentforce Commerce at Salesforce, to explore this paradigm shift. His perspective, shaped by years at the forefront of commerce technology, illuminates why this evolution from search to discovery isn’t just a new feature, but a fundamental change in how brands must engage with their customers to drive growth and loyalty.
From Clicks to Conversation: Aligning with Human Nature
The fundamental problem with traditional search is that it forces humans to behave like machines. We’ve been trained to condense complex needs and desires into two or three keywords, hoping the algorithm on the other side can decipher our intent. This is an unnatural act. The rise of conversational AI has simply reverted us to our default setting: language. As marketing leaders, we must recognize that this shift in consumer behavior isn’t a trend; it’s a correction back to the mean.
Nitin Mangtani frames this evolution not as a technological leap, but as a return to a more natural form of interaction. He sees the past few decades of keyword-based interaction as the anomaly, not the norm.
“The clicks are new to human behavior. Languages and conversations are not new to human behavior. Humans have been conversing for thousands of years, and it’s in fact computers were like a little bit of a, you know, anomaly in that behavior for the last 50 years or so. And so now we are back to how we should all have the most natural way to interact. And same applies to if you own a website or a mobile app.”
This perspective is critical. When we view the challenge through this lens, our approach to site experience changes. We stop thinking about optimizing for keywords and start designing for conversation. Instead of asking, “What query will they type?” we should be asking, “What problem are they trying to solve?” The goal becomes creating an experience that mirrors the best of in-person retail—a knowledgeable associate who listens, understands, and guides. For enterprise brands, this means equipping your digital storefront with an intelligence that can process nuanced requests like, “I need dressier sneakers that go with dark jeans,” a query that would have historically broken most search engines.
The End of “Fire and Forget” Interactions
A core limitation of keyword search is its stateless nature. Each query is an isolated event, a “fire and forget” missile launched into the catalog. The user types “blazers,” gets 200 results, and the interaction is complete. If they want to refine that search, they essentially have to start over with a new query, like “blue blazers” or “casual blazers.” There is no memory, no context, and no sense of a continuing dialogue. This is the antithesis of a helpful, human-centric experience.
The agentic model, by contrast, is built on the concept of a persistent conversation. It mimics the back-and-forth you would have with a store associate, where each new piece of information refines the search in real-time. Mangtani uses the perfect analogy of a skilled salesperson who doesn’t just point you to an aisle, but curates options based on an ongoing dialogue.
“The problem with keyword search was it was fire and forget. There was no follow-on. Versus in real life, there’s never fire and forget. You don’t restart the conversation all over again. You are in a conversation… you continue to have conversation. ‘Like, yeah, actually, I like that blazer. By the way, do you have anything in blue? I wanna try that. Do you have something a little bit more casual?’ So you’re kind of constantly giving the additional prompts to the associate as you are defining what you want, and so you want to mimic that same behavior in the digital world.”
For marketing leaders, this is a profound shift in how we measure engagement. We move from tracking discrete search queries to analyzing the flow and success of entire discovery conversations. Are we helping the customer narrow their choices effectively? Are we anticipating their next question? This conversational context allows for more intelligent upselling and cross-selling, not based on crude rules, but on the natural progression of the customer’s needs within that single, continuous session.
Beyond Keywords to True Intent: The “Taylor Swift Wine” Test
Perhaps the most compelling demonstration of agentic AI’s power is its ability to understand intent that exists outside the strict confines of your product catalog. Traditional search is limited by the data it has: product names, descriptions, and attributes. If a term isn’t in the data, the search fails. This creates a massive blind spot for cultural trends, influencer mentions, and the myriad of ways customers describe products in the real world.
Agentic systems, powered by Large Language Models (LLMs), can tap into a broader universe of knowledge to interpret a user’s intent. Mangtani’s “Taylor Swift wine” example is a brilliant illustration of this. A customer searching for this on a liquor store website would typically hit a dead end, because no such product officially exists. An agentic search understands the cultural context behind the query.
“Let’s say you type in ‘Taylor Swift wine.’ What would you get? Zero search results… We are able to take that knowledge [from social media] and apply this very advanced semantics instead of giving zero search results… You say, ‘Oh, I understand what you mean. Oh, you mean… you’re looking for a white wine, or you like sauvignon blanc, or you like the growers from New Zealand. Let me show you some search results that will come closest to you.’”
This is a game-changer. It transforms a “no results” failure into a successful discovery and a potential sale. For marketers, it means you no longer have to frantically create manual redirects every time a product goes viral on TikTok. The system can connect the dots on its own, translating cultural zeitgeist into commercial opportunity. This capability moves search from a reactive tool that responds to explicit keywords to a proactive engine that interprets implicit intent, capturing revenue from queries that were previously lost forever.
Conclusion: Re-architecting the Digital Relationship
The evolution from search to discovery is more than a feature update; it’s a strategic imperative. It requires us to fundamentally rethink the role of the search box and, by extension, the nature of our digital customer interactions. We are moving from a transactional model, where the user asks and the system fetches, to a relational one, where the brand and the customer engage in a collaborative dialogue to find the perfect solution. The technology is no longer just a utility; it’s a partner in the discovery process, capable of understanding nuance, context, and intent in a way that was previously the sole domain of a brand’s best human associates.
This new paradigm promises not only to lift traditional e-commerce metrics like conversion rates and average order value but also to tackle persistent and costly problems like product returns. By providing more relevant, accurate, and curated results upfront, we can ensure the customer gets what they truly want the first time. As Nitin Mangtani’s vision extends, this intelligence won’t be confined to the website. It will empower store associates on the floor and even syndicate commerce out to the AI channels where customers are beginning their journeys. The brands that embrace this shift won’t just be selling more products; they will be building smarter, more resilient, and more deeply human connections with their customers.
