How to Survive the Era of Agentic Commerce

For 20 years, digital commerce has been built on a single, unshakable assumption: the buyer is a human. We optimised our UX for human eyes; we built "add to cart" buttons for human thumbs; we designed emotional hero banners to trigger human dopamine.

In 2026, that assumption is dead. We have entered the era of agentic commerce. 

The shift we are witnessing today is not just "better search" or "smarter chatbots" – it is fundamental transfer of agency. We are moving from an era where humans search, browse and buy to an era where humans delegate and AI agents execute. The customer is no longer a person; the customer is a protocol.

What is agentic commerce?

Agentic commerce is the ability of an AI system to autonomously navigate the web, interpret commercial intent, negotiate with merchants and execute a transaction on behalf of a user.

It transforms the shopping funnel. In the old world, a user searched for "best running shoes," read three blogs, visited Nike.com and checked out. In the agentic world, a user tells their agent: "Find me running shoes under US$150 that are good for high arches and can be delivered by Friday. Buy the best option."

The agent reads the reviews. The agent checks the inventory. The agent compares the shipping times. And most importantly, the agent presses the buy button.

There is a direct economic impact to each commerce site that you need to be aware of. If you think this is a futuristic concept, look at the numbers from the 2025 holiday season. According to Salesforce data, 20% of global retail sales were already influenced by AI agents, with agent-driven traffic converting at a staggering nine times higher rate than social media referrals.

The shift is accelerating: Morgan Stanley forecasts that agentic shoppers will control US$385bn in US spending by 2030, while McKinsey projects that figure could climb as high as US$1tn. This isn't just about efficiency; it's about survival. Retailers who deployed branded AI agents last year saw sales grow 32% faster than their peers. The wallet share is moving to the machines and it is moving fast.

The new infrastructure: Google’s UCP

For years, the bottleneck to this reality was the "N x N integration problem". Every AI agent (Gemini, ChatGPT, Claude) needed a bespoke API integration with every merchant (Walmart, Target, the corner bakery). It was unscalable.

That changed in January with the release of Google’s Universal Commerce Protocol (UCP).

Found at ucp.dev, this is not just a Google product; it is an open standard co-developed with retail giants like Shopify, Walmart and Wayfair. UCP acts as the USB port for commerce. It allows any merchant to publish a "commerce manifest" – a standardised, machine-readable declaration of their capabilities (inventory, checkout, shipping).

Any UCP-compliant agent can read this manifest and trade with that merchant instantly. It standardises the negotiation. It allows Gemini to say, "I have a user who wants SKU 123. Here is their shipping token. Can you fulfill this?" and allows the merchant to reply, "yes, total is US$140. Payment authorised." 

The major players

The 'agent wars' are no longer about who writes the best poetry. They are about who controls the wallet:

• Google (Gemini): With UCP, Google is aggressively positioning Gemini not just as a search engine, but as a transaction engine. Their new "AI Mode" in Search allows Gemini to execute checkouts directly in the chat interface, keeping the user in the Google ecosystem while the merchant acts as the backend fulfillment node.
• OpenAI (ChatGPT): Through partnerships and their "Operator" capabilities, OpenAI is moving toward a similar model. Their collaboration with Stripe and the Agentic Commerce Protocol (ACP) signals a future where ChatGPT manages your credit card and acts as your primary shopper.
• Anthropic (Claude): Anthropic’s Model Context Protocol (MCP) focuses on allowing agents to read and write to external systems. UCP is designed to be compatible with MCP, meaning Claude can potentially "drive" a web browser to complete complex B2B purchases or fill out procurement forms.
• Perplexity: The dark horse. By partnering directly with payment providers like PayPal, Perplexity is bypassing the traditional ad model entirely, turning answers directly into transactions.

Risks and rewards

This shift is terrifying for some and exhilarating for others. It will honestly make a significant difference in the success of predominantly e-commerce brands. This will be like the early 2000s where, if you don’t create a website, you will either go out of business or you will constantly be behind.

The positives (rewards) include:

• Frictionless revenue: We know that every click reduces conversion by 50%. Agentic commerce reduces the clicks to zero. The "Time to Buy" drops from minutes to milliseconds.
• Objective matching: An agent doesn't care about your hero banner or your font choice. It cares about your data. If you have the best price, the fastest shipping and the right structured data, you win the sale. This democratises commerce for merchants who have great logistics but small marketing budgets.

The negatives (risks) include:

• Brand erosion: If the agent handles the buying, the user never visits your website. They never see your brand story. You risk becoming a "dumb pipe" – a commodity supplier to an AI overlord.
• The "pay to play" bias: Just as we have SEO today, we will have AEO (agent optimisation) tomorrow. There is a high risk that agents will prioritise merchants who pay for "placement" in the agent's consideration set, creating a walled garden.
• The "Amazon defence": We are already seeing major players like Amazon block these agents. The web risks fracturing into "open commerce" (UCP) vs. "walled gardens" (Amazon).

Getting ahead 

So, as a technology leader, what do you do? You stop building for human eyes and start building for machine logic.

To survive in an agentic world, you need an infrastructure that can speak "protocol" fluently. This requires a fundamental shift in how your data is collected, organised and streamed: 

1. Operationalise your data layer: Agents do not read websites; they read data. If your data foundation is messy – if "product view" is named differently in your inventory system versus your analytics – an agent will get confused and move on to a competitor. You must implement a centralised, standardised data layer. This acts as the universal translator for your business, ensuring that every signal (inventory, price, shipping) is defined clearly and consistently in a format that external agents can ingest without error. It must be a data layer that contains clean, consented, contextually-rich data.
2. Shift from batch to stream: Agents operate in milliseconds. If an agent asks, "Is this item in stock?" and your system relies on a nightly batch update, you will sell inventory you don't have, leading to cancellations and being blacklisted by the agent's algorithm. You must move to a real-time data streaming architecture. Your infrastructure needs to be able to broadcast availability, pricing changes and order status updates instantly. In an agentic world, latency is not just an annoyance; it is an invisibility cloak.
3. Bridge the identity gap: This is the hardest challenge. An agent might approach your storefront using a masked, anonymised ID (like a temporary token from Apple or Google). You need an orchestration engine capable of identity stitching. You must be able to recognise that this anonymous agent is actually acting on behalf of your loyal VIP customer, "Sarah." By resolving this identity in real-time, you can authorise the agent to access Sarah’s loyalty points, apply her VIP discount and use her saved shipping preferences, creating a seamless experience that a guest checkout cannot match.
4. Consent is the gatekeeper: Agents are bound by strict privacy protocols. They will often refuse to transact with merchants who do not broadcast clear, compliant consent signals. You need a centralised governance framework that explicitly flags what data can be shared and what tracking is allowed. If your data packet doesn't carry a "consent: true" flag, the agent’s safety rails will likely block the purchase entirely.
5. The bottom line: You cannot retrofit a legacy, batch-based stack for this future. You need a dedicated orchestration layer that sits between your backend systems and these new AI agents – a layer that cleans, structures, enriches with context and streams your commercial data in real-time. The easier you make it for the agent to understand you, the more likely the agent is to buy from you.

Agentic commerce feels disruptive, but it is just the latest evolution. We moved from the high street to the web. We moved from the desktop to mobile. Now, we are moving from the human to the agent.

The merchants who survive will be the ones who realise that their most valuable customer in 2026 isn't the person holding the phone. It's the algorithm running on the server.

Tealium helps companies collect, govern and enrich their customer data in real time to power AI initiatives and improve customer experience.

For more information, visit tealium.com.