Agentic Commerce Means Companies Have to Relearn Their Audiences

More than half of global consumers are using AI to shop, a 70% jump over last year's rates, according to a recent McKinsey report.

Yet while agentic-powered shopping is rapidly evolving – commerce spend is set to hit $1.5 trillion in 2030 – experts are warning that brands and businesses lag behind, sticking to an outdated consumer funnel not designed to target agents.

Businesses, they say, will have to rethink how they engage their customers, and that requires a rethink of traditional business models. Brands need to prepare to communicate with agents, and most companies are experimenting without a clear plan. A Deloitte survey of 200 retail and consumer goods executives found that 40% of consumer goods companies and 29% of retailers have no defined approach to agentic commerce, even though 50% and 60% are already piloting capabilities.

"For twenty years, ecommerce has been designed around one person scrolling one page," Shashank Agrawal, founder of StarApps, which helps brands modernize their product pages to drive conversions, told International Business Times. "Every decision assumed a human was looking: the hero image, the color swatches, the reviews placed just above the fold. That assumption is breaking. More shopping journeys now start with someone typing a request into an AI assistant, and the assistant does the browsing for them."

Leaders recognize the impending importance of agentic AI in reshaping online retail and commerce. What's yet to be established is how they can prepare for that – here is what the strategy should cover, according to experts.

Agents don't read copy; they read data

Here's where much of marketing's philosophy will get dismantled. For years, strategists have based website design and copy on what people can connect to – emotionally, intellectually, spiritually, philosophically, ideologically. A shopping agent isn't moved by any of those sentiments. It doesn't care how punchy and witty a product description is, or how attractive a photo looks. "Agents are built to analyze structured information and metadata: listed attributes, prices, availability, shipping times, and return policies," notes Ruslan Okhrimovych, CEO at Effie.ai.

At a minimum, businesses will need to repurpose their product directories so agents can read them, at a totally different pace than a human review. McKinsey's analysis points to how that will probably look: embedding semantic and behavioral metadata in product catalogs, building interfaces that mean agents authenticate and discover products autonomously, and verifying identity and intent in real time so transactions stay secure.

"An agent compares products, reads specifications, and returns a shortlist. By the time a person lands on a product page, much of the decision has already been made somewhere they never saw," emphasizes Agrawal. Internal data scattered across spreadsheets, disconnected platforms, and siloed departments needs to be unified into a single source of truth.

For most firms, the fix is likely less about collecting new information than restructuring and verifying what already exists. "When we work on a catalogue, roughly half the value comes from turning existing titles, descriptions and specs into clean, structured attributes," said Cristian Orasanu, co-founder of Naratix, a product catalog automation platform. "Agents will only recommend what they can understand and trust, and that's the same data that already powers search, conversion and recommendations today."

Preserving trust while scaling for agents

There are good agents, and there are bad bots. The same technology that paves the way for legitimate agents to shop is also being used to imitate them. According to DataDome, a company that helps manage traffic intent, impersonation of well-known agents is already widespread, with 16.4 million spoofed requests posing as Meta's external agent in 2026 alone.

Companies can't block everything from their websites, however. Indiscriminate blocking has serious consequences for brands, because it can turn away legitimate agent-driven customers. As agents become the front door to commerce, a locked door reads like a closed shop.

Prioritize verification. A bot cannot be trusted simply on what it says it is, so businesses need mechanisms to confirm who an agent is, who authorized it, and what it is allowed to do – for every single agent visiting their website.

Albert Castellana Lluís, co-founder and CEO of GenLayer Labs, a company building trust infrastructure for AI agents and the world's first digital court to resolve agent disputes, said businesses should make their promises easy to check: what's being sold, the full price, delivery terms, and what happens if something goes wrong. "Keep a record of what the agent agreed to and what you delivered. A verified identity tells you who you're dealing with. It doesn't tell you whether either side kept the deal."

Two audiences, one customer, and the metrics that matter

Businesses are selling to two audiences: the agent that evaluates and the person who instructs it. That individual might never visit the site or be exposed to an ad, so trust and targeting are built in two places. The agent needs reliable, structured, verifiable data. The person needs a reason to tell their agents to prefer a specific brand over another.

Persuasion and brand equity change. Much of the marketer's toolkit, including aspirational storytelling and urgency-driven copy (usually built on FOMO), has been designed to move a human in a moment. An agent comparing hundreds, even thousands of options on a customer's behalf in a given moment won't feel any of this. But the brand doesn't disappear; it changes form. An agent won't care that a certain celebrity endorses a brand's handbag, but it will quantify the average of five-star reviews and volume, in addition to return rates, delivery reliability, warranty terms, and refund policies. Brands can survive if their reputation is also built on what machines can verify, not just sentimental value to connect with people.

Businesses will also need to change their approach to measuring performance. Click-through rates, impressions, and conversion rates are metrics built on human intent and engagement. An agent doesn't click an ad. Leaders should consider tracking how often products are perused and selected by agents. One option is LLM tracking tools. They also need to keep track of how much agent traffic is verified rather than spoofed, and how agents' involvement impacts revenue. So, the focus moves beyond the attention and engagement people give to the work agents do.

What to do now

The first move should be to audit product data and find out how agents are already reaching your website. That gives leaders visibility into what their strategy actually needs before moving ahead with a pilot that outpaces a blueprint.

Effie AI also strongly urges businesses to set clear rules for what an agent can and can't do on their site, and connect those rules to live inventory, pricing, and policies. "Governance without integration delivers nothing, and integration without governance creates risk," added Okhrimovych.

Marketing will always matter. Businesses now need to include agents in their strategies, too, and lay the groundwork to welcome a new kind of shopper to their websites.