The narrative of digital transformation in retail has been remarkably consistent. First, get online. Then build an app. Then optimize checkout. Add loyalty points, push notifications, maybe a chatbot.
For years, this was the right strategy. It moved retail from physical-only to omnichannel, giving customers access, convenience, and choice.
But something fundamental shifted at Google I/O 2026.
The conversation is no longer about getting customers to your storefront—digital or physical. The question retailers must now answer is: When your customer's agent comes shopping, will your business be ready?
This is the move from HTTP/HTTPS (where humans browse sites) to the Universal Commerce Protocol (UCP) (where agents act across platforms). It redefines not just the technology of retail, but the entire relationship between businesses, customers, and shopping itself.
The Technology: Where HTTPS Met Its Match
To understand what's happening, it helps to think about HTTP's role in web history. HTTP standardized how browsers and servers communicate, unlocking the web as we know it. Similarly, UCP is designed to standardize how agents—autonomous systems acting on behalf of customers—can understand and transact across different retailers, payment systems, inventory platforms, and fulfillment networks.
But UCP isn't just a protocol upgrade. It's accompanied by three critical technical announcements that fundamentally change what's possible:
1. Gemini Spark and Proactive Agent Commerce
Spark is a browser-based agent that doesn't wait for you to ask. It reads your calendar, understands your commitments, and anticipates your needs. This is where shopping transforms from a task you remember into a background process that simply happens.
For someone like me—juggling data science work, team leadership, community projects, and a dozen other commitments—Spark represents a genuine return of time. When I see on my calendar that I'm on "snack duty" for a weekend game, the agent can:
- Understand the context (when, where, how many people)
- Recall my constraints (dietary preferences, budget ranges, previous choices)
- Proactively set up an Instacart order
- Remember to exclude nuts because someone in the group has allergies
- Present the order for approval before checkout
This isn't a recommendation. This is delegation.
2. Agent Payments Protocol (AP2) and Trust Through Guardrails
With great delegation comes the need for security. AP2 allows customers to set strict, tamper-proof guardrails that agents must follow:
- Spending caps per transaction
- Brand whitelists and blacklists
- Category restrictions
- Approval thresholds
Every transaction creates a cryptographically secure digital paper trail. The customer maintains full accountability and can revoke agent access instantly.
This solves what traditional fintech couldn't: How do you let an autonomous system handle money on your behalf while maintaining control?
3. Gemini Nano and Offline-First Shopping
The final piece is perhaps the most underrated: Gemini Nano 4, an on-device model powerful enough to run complex reasoning entirely locally—in airplane mode, with no internet connection.
This matters because shopping doesn't always happen online. Sometimes you're in a mall with spotty WiFi, trying to make style decisions for formal wear, and you need product information, sizing guides, or virtual try-on capabilities without relying on the network.
With Nano, a custom retail app can:
- Store product catalogues locally
- Run visual recognition for style matching
- Provide detailed product comparisons
- Offer sizing guidance
- Create virtual try-ons
All without a data connection. The technology disappears, and you get back to what actually matters: the decision.
Why This Matters Personally (And Why It's Bigger Than Me)
For years, I've approached shopping the same way I approach most of life: pragmatically. I don't have time for multiple apps, price comparisons, or decision paralysis. I have outcomes I need.
The problem is that retail was built assuming everyone wants to browse, compare, and feel in control of the process. But that's not universal. People shop differently based on their constraints.
For the time-poor (me): The best shopping experience isn't browsing. It's delegation. Spark handles the thinking; I handle the approval.
For the efficiency-focused: It's not about the most beautiful app. It's about an agent that understands your priorities—best deals, minimal time, trusted brands—and orchestrates the entire journey. Tools like Antigravity 2.0, Google's agent-first development platform, allow retailers to build custom agents that act as mission control. These agents can map optimal routes across stores, identify real-time deals, research product quality, and even execute purchases under strict conditions using AP2. For someone who refuses to waste time but won't sacrifice value, this is transformative.
For the expert shoppers (like my cousin with formal wear): It's about augmentation. Having an on-device assistant that knows her style preferences, past purchases, and body measurements while she's physically shopping—even offline—amplifies her decision-making without replacing it.
The retail insight here is stark: Customers aren't losing interest in shopping. They're losing patience for friction. They're not abandoning carts because the prices are too high; they're abandoning because the decision-making burden is too heavy, the process too fragmented, the experience too disconnected from how they actually live.
The Business Case: Capturing the Opportunity Cost
Most retailers think about lost sales in terms of price, stock, marketing budget, or abandoned carts.
But there's another category that deserves attention: sales lost because shopping didn't fit into the customer's life.
A customer doesn't buy because they're busy. Another delays because they're overwhelmed by options. Another avoids the store because the trip feels exhausting. Another abandons the cart because comparing offers is too much work.
This is opportunity cost—and it's enormous.
Agentic retail has the potential to recover this by helping customers shop around their actual constraints, not just their preferences.
The business question shifts from: "What does this customer want?"
To: "What is making this purchase harder than it needs to be?"
Why Retailers Need to Get Agent-Ready Now
Here's where the strategic opportunity lies for retailers, especially those looking to compete globally.
UCP is open-source and protocol-based, which means it's not controlled by any single platform. But adoption requires retailers to think beyond their website or app. The backend systems—product data, inventory accuracy, pricing consistency, return policies, loyalty rules, payment flows—must be reliable enough for agents to understand and trust.
A traditional digital transformation question: "How do we get customers to our website or app?"
An agentic commerce question: "Are our systems clean, accurate, and trustworthy enough for an autonomous agent to recommend us?"
This sounds subtle, but it's profound. A beautiful website helps humans decide. Clean, consistent, machine-readable data helps agents recommend you.
The Universal Cart, which works across Search, YouTube, and Gmail, is an example of this in action. It can even apply intelligent reasoning—flagging when a processor selected in one cart isn't compatible with a motherboard in another, catching errors before purchase.
For retailers, this means:
- Invest in data infrastructure. Product catalogues, stock positions, pricing feeds, and policy documentation must be accurate and machine-readable.
- Think about agent-first UX. Not every customer needs a beautiful storefront anymore; some need to be easily found and trusted by an agent.
- Build for transparency. Agents will explain why they recommended your product over a competitor's. Can your systems support that narrative?
The Intelligence Question—And Why It Matters for Trust
Here's where this gets complex.
Traditional retail systems mostly see what you bought. An agentic system understands what you intended to buy, what you compared, what you rejected, what mattered most, and what constraint shaped the final decision.
That intelligence is valuable. It could improve demand planning, reduce waste, enable better promotions, and help retailers understand market behavior with real nuance.
But it also raises uncomfortable questions:
- Who owns intent data? If an agent learns that you're willing to spend more on sustainable products, or that you're price-sensitive, who controls that insight?
- What does real consent look like? A checkbox saying "I agree to share shopping data" isn't consent; it's legal cover.
- Can you see what your agent shared? Transparency has to flow both ways.
- Can you revoke access? Consent must be revocable, not permanent.
- Why was this recommended over that? The agent's reasoning should be explainable.
This isn't just a technical problem to solve; it's a governance conversation.
The retailers who build agentic commerce responsibly—with transparent data handling, genuine consent, and customer control—will win trust. Those who treat it as an opportunity to extract data quietly in the background will eventually face backlash.
The strongest competitive advantage in agentic retail won't come from having the best algorithms. It will come from being the brand customers trust their agents with.
What Comes Next
The shift from HTTP to UCP represents more than a technology transition. It's a redefinition of what retail is for.
Shopping in the age of agents isn't about having more options. It's about removing the friction that keeps people from getting what they need.
For me, that means delegating the thinking.
For others, it means efficiency.
For some, it means confidence in a decision.
The retailer—and the technology platform—that understands this will thrive.
But understanding it means asking better questions:
- What's making this purchase harder than it needs to be?
- Are our systems ready for agents to trust us?
- Are we building this responsibly?
These are the questions that will separate retail leaders from everyone else in the agentic era.
The question is no longer: "Does your retail business have an app?"
The question is: When your customer's agent comes shopping, will your business be ready?
And more importantly: Will your customer trust you with their agent?
This is a submission for the Google I/O Writing Challenge
Top comments (0)