Agentic Commerce: What It Means for Your Business

The next customer journey may not begin on your website. A buyer might ask an AI assistant which product fits a particular need, question a brand through an ad, and move toward checkout without ever following the familiar sequence of search results, website pages and retargeting.

That does not mean the funnel has disappeared. It means the funnel is changing shape — and the gaps in it are about to get more expensive.

For more than forty years I have helped businesses connect marketing activity to growth, profit and brand strength. The current wave of AI agents reinforces a lesson I have seen repeatedly: new technology can accelerate a good strategy, but it cannot compensate for unclear positioning, unreliable information, or a poor customer experience.

will AI agents start buying from my business?Possibly. But a recommendation only has commercial value if the buyer can understand your offer, trust you, and complete the next step — and that is a business problem, not an AI problem.

What actually changed in September 2026

Four developments, inside four weeks, show how quickly discovery and conversion are converging. I am listing them with sources because every one of them is checkable, and a claim about this field that cannot be checked is worth nothing.

ChatGPT Ads came to Shopify. Shopify merchants in the US can now connect their catalogue and commerce-event data to ChatGPT Ads, build product campaigns against live inventory, and measure results in the Analytics tab or in ChatGPT Ads Manager. This is materially different from hoping a product turns up in an AI recommendation. It is a paid channel that can be tested against a business outcome.

Google put a conversational agent inside YouTube ads. September's Demand Gen update introduced a Business Agent experience, letting viewers ask questions about products or brands alongside a video ad, plus one-tap paths from image ads on YouTube Shorts and Gmail. The implication is simple: an ad is becoming an interactive part of the buying conversation rather than a message that points somewhere else.

Shopify's agentic storefronts went multi-channel. Eligible merchants' products are now discoverable through ChatGPT, Google AI Mode and Gemini, Microsoft Copilot, and Meta. The checkout differs by channel in a way that matters for planning: ChatGPT sends the buyer to complete the purchase on the merchant's own online-store checkout, in an in-app browser or a new tab, while the other three channels can offer Shopify-powered direct checkout without the customer leaving the AI surface.

That distinction is not a technicality. On one channel you own the final screen — your branding, your upsell, your analytics. On the others you do not. Decide which you are optimising for before you read a single performance report.

Six banks published shared principles for agentic commerce. On 22 September, ASB Bank, Bank of America, Capital One, Commonwealth Bank of Australia, ING and NatWest jointly set out five areas for trusted agentic commerce: transparency, safety, privacy and data, choice, and interoperability. The stated aim is that customers and merchants retain choice, control and flexibility as agents begin transacting on people's behalf.

These are meaningful changes. They are not proof that every business should shift its budget to AI advertising tomorrow, or that autonomous agents will become anyone's largest sales channel this year. They are a reason to get ready and to test carefully.

Visibility is only the first step

Many companies are asking, “will AI recommend us?” That is a useful question and it is not a sufficient one. A recommendation has commercial value only if the buyer can understand the offer, trust the business, and complete the next step.

Consider an e-commerce brand with incomplete product descriptions, inconsistent prices, unclear shipping terms or stale stock data. An assistant may well surface the product. The buyer's next three questions will expose every one of those weaknesses, faster than a human salesperson ever would. For a professional-services firm the equivalent gaps are vague positioning, unsupported claims, no visible proof of expertise, and no obvious way to book the right conversation.

This is why I advise clients to work backward from the customer decision rather than forward from the technology. What does the buyer need to know? What objections will they raise? Which claims can you actually substantiate? What should happen the moment interest turns into intent?

The goal is not to optimise for AI in isolation. It is to build a business that is easy for both people and machines to understand — and easy for customers to trust. Those turn out to be the same project.

Trust is becoming part of the conversion system

As agents move closer to transacting on a customer's behalf, control and accountability stop being abstract. Customers need to know what they are agreeing to, whether an agent is acting for them, what happens if it makes a mistake, and how returns or disputes get handled. A smooth automated experience that creates uncertainty will damage both conversion and brand equity, and it will do it quietly.

My recommendation is to keep human approval in place for consequential actions until the rules, permissions and exception paths are clear. Automation should reduce friction without removing responsibility. Those are not the same thing, and the businesses that confuse them will find out expensively.

Measure what actually changed

Here is the lesson from this month that almost everyone will miss. Between 21 and 23 September, Shopify changed how it measures sessions: sessions now continue based on activity rather than resetting at midnight UTC, sessions without pageviews are counted, and identified bot sessions are filtered out by default.

Shopify's own guidance is unusually direct about what this means. A change in these metrics does not necessarily mean your traffic, orders or customer behaviour changed — it reflects an update to how session activity is measured. A higher or lower conversion rate after the update is not automatically good or bad.

0actual change in customer behaviour required for a business to see its conversion rate move after 21 September. Orders, sales and customer counts were not affected by the update.

So a business could celebrate a higher conversion rate, or panic about lower traffic, entirely for the wrong reason. Before declaring any AI-channel pilot a success, confirm tracking is working, note any measurement changes, and compare orders, revenue, margin and new-customer quality alongside traffic and conversion rate.

The same discipline applies to every new AI tool. Define the baseline first. Change one thing at a time where you can. Judge the result by commercial value rather than activity.

A practical starting plan

For professionals and e-commerce leaders, I would start with four questions:

Can buyers and AI systems clearly understand your offer? Review your services or product catalogue, your differentiators, pricing context, proof, policies, and the next step you want someone to take.
Where does your funnel lose value today? Find the points where qualified interest stalls — slow responses, unanswered objections, thin product detail, abandoned carts, weak follow-up.
What can an agent safely improve? Choose one bounded use case: answering routine product questions, qualifying inquiries, helping a customer find the right item. Define precisely when a person takes over.
What result will justify the investment? Set one measurable target — better qualified leads, improved conversion, reduced service time, higher repeat purchase, or stronger profit per customer.

AI is opening real new doors to discovery and commerce. But the businesses that benefit most will not be the ones that deploy the most agents. They will be the ones that combine clear strategy, trustworthy information, disciplined measurement and a genuinely better customer experience.

That is how technology turns attention into durable business value. It is the same way it always has.

Start with whether you are visible at all. The free audit asks a live assistant real customer questions and shows whether your business comes back — and which sources it used to decide.

Run the free audit

Common questions

What is agentic commerce?
Agentic commerce is where an AI assistant helps a customer discover, evaluate and in some cases purchase a product, rather than sending them to a website to do it themselves. In practice it spans a range: an assistant recommending a product the customer then buys on your site, through to an agent completing a transaction on the customer's behalf. As of September 2026 the first is common and the second is still emerging.

Will AI agents replace my website?
No, and the September changes point the other way. Shopify's agentic storefronts send ChatGPT buyers to the merchant's own checkout to complete the purchase. Your website remains where the transaction finishes, where your branding lives and where your analytics work. What is changing is that it may no longer be where the journey starts, so the site has to answer questions a buyer has already half-answered elsewhere.

Should I advertise inside ChatGPT?
If you are already named in AI answers, it is worth a test budget rather than a strategy: real intent, thin competition, immature measurement. If you are invisible in AI answers, fix that first. Buying the slot beneath an answer that names your competitor is an expensive way to look like the alternative.

How do I know if an AI channel pilot actually worked?
Define the baseline before you start, then judge it on orders, revenue, margin and new-customer quality rather than sessions and conversion rate alone. Platform measurement changes independently of your performance — Shopify's September session update moved conversion rates without any change in customer behaviour — so confirm what your tracking is doing before you attribute a result to the channel.

What should I do first if I am not ready for any of this?
Make your offer unambiguous to a machine and to a person: complete product or service detail, consistent pricing, clear policies, substantiated claims, and one obvious next step. Every agentic channel reads the same underlying facts about your business. Fixing them once improves every channel at the same time, which is why it is the only part of this worth doing before you know which channels matter.

Sources

Every factual claim above links to its primary source, because a claim about a fast-moving field that you cannot check yourself is not worth much:

About the author. Siamak Kalhor is an AI advisor and marketing strategist with more than forty years of experience helping businesses improve their marketing systems, close funnel gaps, strengthen their brands and pursue sustainable growth. At Siamak Kalhor Consulting he combines human judgment with practical AI transformation for professionals and e-commerce leaders. Discuss an AI and marketing systems assessment.

AI can recommend you. It still cannot close the sale.