AI

The now and next steps retailers should take in the agentic commerce era

The projected promise and perils of agentic commerce seem to change almost daily, and that can be disconcerting. A thoughtful approach now will put merchants on the path toward stability and future success.

Adobe Stock

June 30, 2026 by Nicole Jass

Summarize with

Agentic commerce is moving both faster and slower than the shopping technology's significant buzz would indicate.

Agentic is exploding as a revolutionary discovery tool for consumers: Help me decide between these two products; is this product really worth more money; what can help me sleep better without side effects, etc. But as a set-it-and-forget it, wake-me-when-my-product-arrives helpmate, the agentic hype is outrunning reality.

The data tells the story: While the number of online orders resulting from LLM referrals is exploding from a small base — up over 600% year over year on Signifyd's Commerce Network — the total number of purchases executed by AI agents remains at fewer than 1% of all orders.

That aligns with what we are hearing from customers, who say they are not seeing significant conversions from AI agents. It also reflects the reality of moves by agentic commerce platforms like OpenAI, which recently rolled back its Instant Checkout feature, acknowledging that consumers are using the technology for browsing, but not buying.

That will change, and merchants know they need to be ready. Rarely is there a crawl-walk-run situation in e-commerce. It's a "run-run-run" business. Even so, online brands need to be thoughtful about their approach to agentic.

Perhaps what retailers need in order to be most effective is a measured, phased mindset — in other words a "jog-run" mindset.

What can retailers focus on now to embrace the agentic commerce era

Now is the time to cast an agentic eye on your customer-facing content. That's the "run" phase.

Data needs to be structured today to appeal to autonomous agents. It must be accurate and consistent, so bots can ingest prices and delivery times. Product catalogs need to be up to date, so that stated inventory reflects reality. Return policies need to be crisp, coherent, comprehensive, and consistent.

The biggest danger for merchants in the agentic era is not falling victim to bad bots and fraud schemes; it's turning away good bots that are fast becoming the gateway to product pages and will ultimately deliver a significant portion of conversions at checkout. Think of it as a new form of headless commerce that you don't quite control, but you sure can curate through thoughtful and accurate data availability.

While e-commerce and Gen AI platforms are finding their footing when it comes to the alphabet soup of protocols and agentic tools — ACP, UCP, A2A, MCP — merchants need to consider the novel fraud and risk vulnerabilities this shift in shopping introduces. This is the "jog" phase.

How agentic commerce challenges traditional signals and approaches to fraud

Agentic commerce shifts consumers' entry point to a shopping journey far upstream from a merchant's digital site. That's the key disruption to traditional fraud screening and those signals change in several ways:

Traditional online orders arrive with an IP address, which provides helpful location confirmation. Orders from AI agents generally include the IP address of a large data center in the office-park-dotted exurbs.
With many traditional orders, a merchant can review past transaction history for insights into associated risk. AI agents are not known users. Their history is a clean slate.
Human customers provide stable device ID with a discernible history. Agents do not. Their device IDs are ephemeral, spun up in the moment.
Humans browse, add to cart, remove items from cart, browse some more, giving clues to their intent. Agents act via APIs and transact in milliseconds, providing no behavioral clues.
An agent's email address will reflect its platform. Human orders come with recognizable, often established, email addresses.
Human shoppers often create accounts for convenience and to take advantage of loyalty programs. Bots have no such concerns, so no accounts or programs are needed.

Four ways to prepare for the future of agentic checkout

All this means merchants need to think about agentic orders and risk differently. The slower-than-expected rollout of widespread agentic checkout gives them that opportunity. Here are the keys to preparing for the era of end-to-end agentic commerce:

Shake the notion that all bots are bad. Some may be among your best customers. Invest in building or updating machine learning models that are able to detect the differences among human shoppers, scraper bots, and beneficial agent traffic.
Identify and authenticate the agentic commerce platforms — for instance, OpenAI, Perplexity and Gemini — that are behind visits and transactions using cryptographic checks.
Rely on a layered approach, authenticating the agent, and authenticating the human behind the agent. Designated authorization supported by cryptographic key binding provides a verifiable link between the agent and the human it represents.
Rely on a broad cross-merchant network to access the intelligence needed to link an agent back to its human client's network history. That provides necessary insights to establish the agent and its client as trustworthy customers.