At BrightonSEO San Diego, Forrester analyst Nikhil Lai offered a useful counterweight to much of the excitement surrounding AI-powered shopping: Agentic commerce is not inevitable.

The technology is advancing quickly. AI agents can already research products, compare options, and move between systems in pursuit of a goal. But technical capability alone will not persuade consumers to surrender control of a purchase or convince retailers to reorganize their operations around a new sales channel.

For agentic commerce to move from an intriguing experiment to a meaningful part of ecommerce, several things must work at once. The product selection must be good, the recommendations must be trustworthy, checkout must be nearly effortless, product and operational data must be complete, and the organization behind it all must be able to adapt.

That is a much higher bar than simply adding a shopping assistant to a website.

What Is Agentic Commerce?

Lai defined agentic AI as technology that acts autonomously across systems to achieve a given goal.

In commerce, that could mean asking an AI assistant to find a product that meets a set of requirements, compare the available choices, select one, and potentially complete the purchase. Instead of moving manually from search engine to retailer to review site to checkout, the consumer delegates some or all of that journey to an agent.

The distinction matters. A chatbot may answer a question. An agent takes action.

But the ability to take action does not guarantee that consumers will want agents to act or that the available commerce ecosystem will produce good results.

The Weakness: Adverse Selection

One of Lai's most important cautions was the risk of adverse selection.

An AI shopping engine can recommend only from the products, merchants, and information available to it. If the strongest retailers do not participate, or if commercial incentives cause certain products to receive preferred placement, the agent's inventory may not represent the best options for the customer.

That creates a fundamental trust problem. Consumers will not continue delegating purchases to an agent if they suspect it is choosing from an inferior pool of products or favoring whatever a platform has been paid to promote.

This is one reason Lai resisted treating agentic commerce as a foregone conclusion. The technology may work perfectly while the market surrounding it produces recommendations consumers do not trust.

The Opportunity: Apple Pay-Like Simplicity

The clearest opportunity is an Apple Pay-like experience inside an answer engine: a familiar, low-friction payment process that lets a customer move from recommendation to purchase without repeatedly entering shipping and payment information.

Reducing checkout friction could make AI-assisted purchasing much more attractive. It would also move companies such as OpenAI, Google, and Anthropic closer to the transaction itself.

That creates a new problem. The closer an answer engine gets to the point of purchase, the more responsibility it may assume for everything that happens afterward. Who handles a chargeback? Who resolves a dispute? What happens when an item is late, damaged, counterfeit, or simply not what the customer expected?

Those functions are far removed from the core businesses of most AI companies. Yet failures at that stage would not feel separate to the customer. A bad delivery or unresolved return could damage trust in the agent that recommended and facilitated the purchase.

Agentic commerce therefore requires more than a good recommendation and an elegant checkout. It requires a credible post-purchase experience.

Where Agentic Commerce May Fit First

Lai suggested that the strongest initial opportunities may exist at opposite ends of the consideration spectrum.

Low-consideration products, such as paper towels or other routine purchases, are natural candidates for direct agent purchasing. The customer knows roughly what is needed, the risk is limited, and convenience may matter more than extensive comparison.

High-consideration products, such as consumer electronics, are likely to generate substantial AI-assisted research. Consumers may spend 20 or 30 minutes asking detailed questions, comparing features, refining requirements, and evaluating tradeoffs before making a decision.

Products in the middle may be less natural fits. They may not be routine enough to delegate completely, but they may not warrant the extended research session that makes an answer engine especially valuable.

This means brands should not begin with the question, "How do we participate in agentic commerce?" They should first ask, "What role could an agent realistically play in the purchase of our products?"

Forrester's Agentic Commerce Framework

Lai presented a four-part framework for evaluating readiness.

1. Agentic Fit
   • Organizations must assess how well their products match agent-assisted behavior. Is the purchase routine enough to automate? Is it complex enough to benefit from extended research? What information and reassurance does the customer require before acting?
   • The objective is not to force every product into an agentic journey. It is to identify where an agent genuinely reduces effort or improves a decision.
2. Machine Advantage
   • An AI system is only as useful as the information it can access and understand. Brands need rich, accurate content that gives machines an advantage in answering customer questions.
   • That includes more than polished marketing copy. Machines need concrete details such as materials, dimensions, compatibility, features, certifications, use cases, and other attributes that help them compare products and match them to specific needs.
   • If the information is missing or ambiguous, the product may be omitted from the answer or represented incorrectly.
3. Commerce Operations
   • Product information must be supported by reliable operational data. Assortment, inventory, pricing, availability, and fulfillment details all need to be complete and current.
   • An agent cannot deliver a trustworthy shopping experience if it recommends an unavailable product, quotes an outdated price, or promises delivery that the merchant cannot meet. In agentic commerce, back-end accuracy becomes part of the customer experience.
4. Organizational Adaptability
   • Lai described organizational adaptability as perhaps the greatest challenge.
   • Agentic commerce crosses traditional boundaries. Digital business, ecommerce, IT, content, SEO, paid search, social, public relations, legal, and information security may all have a role. If those teams work toward different goals, even excellent technology and data will not produce a coherent experience.
   • Digital business and IT are a particularly important partnership. Lai recommended establishing shared objectives, for example, a target for sales originating in an answer engine and fulfilled on the retailer's site, so the teams are working toward the same outcome.

Listen to What Customers Ask AI

For marketers, one of the most immediate opportunities is to study the prompts consumers use when researching products.

Lai said Forrester's consumer data shows that shoppers may ask as many as eight follow-up questions after an initial prompt. Individual questions may exceed 20 words, and an entire research session can last 20 to 30 minutes.

That represents an unusually rich source of voice-of-customer data. These are not two- or three-word search queries. They are detailed expressions of needs, constraints, concerns, preferences, and commercial intent.

Marketers can use those questions to shape comparison guides, FAQs, articles, videos, infographics, product-page content, and other resources. Listening to the conversation is how a brand learns what it must explain in order to appear and remain useful during AI-assisted research.

Make Content Accessible to Answer Engines

Creating the right content is useful only if AI systems can retrieve it.

Lai recommended limiting excessive JavaScript on pages that brands want answer-engine crawlers to index. Google's crawler is highly sophisticated, but other crawlers may struggle to process pages that depend heavily on JavaScript.

He also highlighted IndexNow, a protocol that allows websites to notify participating search engines when content has been added, changed, or removed. This matters because many answer engines rely heavily on Bing's index. Making updates easier to discover increases the likelihood that current information will be available when consumers ask relevant questions.

Structured data is equally important. Author schema can help establish who created a piece of content and that person's expertise. Organization schema can help machines understand relationships between a parent company and its brands. Product schema and detailed attributes help answer engines identify exactly what is being sold.

The underlying principle is simple: Content must be useful to people and legible to machines.

Use AI for Content Velocity With Human Oversight

Agentic commerce may increase the amount and variety of content companies need. AI can help teams brainstorm, draft, repurpose, and scale that work.

But Lai offered a clear warning: Unedited AI content is a failure.

AI may help a company keep pace with rising content demands, but publishing generic machine-generated material simply creates more noise. Human judgment is still needed to verify claims, add expertise, preserve the brand's point of view, and ensure the content genuinely answers a customer's question.

The goal is not maximum output. It is useful coverage at a sustainable scale.

Search Marketing Must Become More Holistic

As more commercial intent moves into answer engines, the boundaries among SEO, paid search, and emerging AI advertising will become less meaningful.

Lai described the shift as moving toward holistic search or total search: conventional SEO combined with paid search and the next generation of answer-engine marketing.

That requires organizations to:

• Test whether paid placements are generating incremental value rather than taking credit for purchases that would have happened organically.
• Bring paid and organic performance data into a unified view.
• Develop measurements that show how channels influence one another, such as organic-assisted paid-search revenue.
• Watch for new advertising and commerce opportunities inside AI interfaces.

This is not simply an expansion of SEO. It is a reorganization of search around the places where people now express intent.

Replace Channel Silos With Flexible Budgets

The same logic applies to media spending.

Rigid budgets divided among SEO, paid search, social, retail media, and other channels make it difficult to respond to a rapidly changing environment. Lai said leading brands are beginning to think in terms of a more fluid pool of investment that can be moved toward promising opportunities.

That flexibility allows companies to test emerging formats, hold them accountable to appropriate performance measures, and increase investment quickly when something works. It also prevents teams from protecting a channel budget at the expense of the broader customer journey.

Prepare Without Assuming

The most useful aspect of Lai's presentation was its balance. Agentic commerce could transform how people research and purchase products, but transformation is not guaranteed simply because the technology exists.

Brands should prepare now by improving content, product data, operational accuracy, technical accessibility, measurement, and cross-functional alignment. Those investments will make them more understandable and trustworthy across conventional search, answer engines, and future shopping agents.

But they should also remain clear-eyed. Agentic commerce will succeed only if it produces choices consumers trust, payments that feel effortless, fulfillment that meets expectations, and customer service that works when something goes wrong.

The future of shopping may be agentic. The companies most likely to benefit will be the ones that treat it not as inevitable, but as something they must earn.