Ulta Beauty’s Agentic Commerce Bet: Gemini Meets a 46 Million–Strong Loyalty Engine
Ulta Beauty is turning its vast loyalty ecosystem into a testbed for agentic commerce. The US-based specialty beauty retailer is rolling out two Gemini-powered experiences: shoppable products within Google’s AI Mode in Search and the Gemini app, and Ulta AI, a new AI shopping assistant embedded in its owned digital channels. In Google’s conversational interfaces, shoppers can ask for Ulta Beauty product recommendations, compare options, and complete a streamlined checkout journey, all backed by the Universal Commerce Protocol, an open standard for agentic commerce. On its own properties, Ulta AI is built on Gemini Enterprise for Customer Experience and taps into insights from 46 million loyalty members, who account for 95 per cent of sales. That first-party data underpins Ulta’s ambition to be present across the entire beauty journey and to elevate relevance and impact in every interaction.
From Chatbots to Agents: How Agentic Commerce Actually Works
Ulta Beauty’s latest rollout illustrates the shift from simple AI chatbots to agentic commerce. Traditional AI shopping assistants mostly answer questions or search product catalogues. Agentic systems go further: they actively guide the shopping journey, propose curated paths, and can execute parts of the transaction. Instead of merely surfacing a list of foundations, an AI shopping assistant can ask clarifying questions about skin type, coverage preferences, or previous purchases, then assemble a personalised short list, suggest complementary items, and move the shopper smoothly to checkout. The distinction mirrors a broader AI challenge already evident in supply-chain technology: visibility is not the same as decision-making. In logistics, knowing a shipment is delayed does not automatically trigger the right response without embedded decision logic and authority. Similarly, in commerce, the value lies in AI agents that can weigh context, apply business rules, and help complete actions, not just generate more product “visibility.”
CMO–CTO Fusion: Why Org Design Matters for Retail Personalisation
Ulta Beauty’s agentic commerce push is not just a technology story; it is an organisational one. The company highlights a tight partnership between marketing and technology leadership as a catalyst for faster experimentation and deployment. Having already laid down an Adobe-based personalisation stack, including Real-Time CDP and orchestration, Ulta can now plug Gemini-powered agents into an existing framework of audience segmentation, content, and testing. This CMO–CTO “functional fusion” lets brand, data, and engineering teams align on a single North Star: relevance and impact for guests. It reduces the friction that often bogs down AI initiatives, such as scattered decision rules or unclear execution authority. In effect, it helps close the gap between insight and action that many organisations struggle with in other domains, where AI can detect issues or opportunities but still relies on manual workflows to actually change the customer or operational outcome.
What Shoppers Gain: Smarter Discovery, Bundles and Beauty-Goal Guidance
For shoppers, Ulta’s AI shopping assistant promises more than faster search. By drawing on preferences, dislikes, and past purchases from its loyalty base, Ulta AI is designed to surface highly relevant product suggestions across 600-plus brands. Agentic commerce allows the assistant to understand broader beauty goals—such as building a minimal skincare routine or finding a look for a specific event—rather than just matching product names. It can assemble smarter bundles, recommend appropriate substitutes, and help navigate large catalogues without forcing users to know the right filters or jargon. As consumers increasingly use AI to discover, evaluate, and even complete shopping journeys, this kind of guided, conversational experience could become the new default. If executed well, the result is less trial-and-error, fewer abandoned carts, and a more continuous, post-purchase relationship where the assistant can also support use, replenishment, and next-step recommendations.
Implications for Global Retailers, from Fashion to Southeast Asian Marketplaces
Ulta Beauty’s move signals how agentic commerce could spread across e-commerce categories and regions. To make similar AI assistants work, retailers must tightly integrate product data, inventory, promotions, and customer profiles so agents do not hallucinate or recommend out-of-stock items. Decision logic—about discounts, substitutions, and service priorities—needs to be encoded, not scattered across spreadsheets and siloed teams, or the AI will remain a passive advisor. Fashion players could use agents to build outfits and capsules around lifestyle cues; electronics retailers could guide complex, spec-heavy purchases. Southeast Asian marketplaces, with highly diverse assortments and price-sensitive shoppers, can adapt agentic models to local languages, payment methods, and social commerce habits, using first-party data from loyalty or app ecosystems. Across all markets, guardrails will be critical: monitoring bias in recommendations, ensuring transparent data use, and preserving human override for sensitive, high-value, or regulated purchase decisions.