A shopper opens ChatGPT and types "best magnesium for sleep, ships to the UK". The agent reads the reviews, weighs up three brands, and offers to buy the winner right there in the chat. Your product page, your carefully built funnel, the ad you paid for. None of it got a look in.
This is not a thought experiment any more. Traffic to US retail sites from generative AI sources jumped 693% year on year over the 2025 holiday season, and shoppers arriving from those AI sources converted 31% more than shoppers from other channels, according to Adobe Analytics. Salesforce put a number on the whole season: AI and agents influenced $262 billion in online holiday sales, close to a fifth of everything sold, per Salesforce.
So the worry is fair. If an agent picks the product and completes the sale, does the discovery game you have been playing still pay? My answer is that agentic commerce does not kill retention. It makes retention the only ground you fully own.
What is agentic commerce, in plain terms?
Agentic commerce is when an AI agent does the shopping for a person: it searches, compares, and can complete the checkout on their behalf. The person sets the intent, the agent handles the legwork, and the purchase can finish inside the chat instead of on your website.
This got real fast because the rails now exist. OpenAI and Stripe shipped Instant Checkout in ChatGPT with an open Agentic Commerce Protocol, live with Etsy sellers and rolling out to over a million Shopify merchants, per Stripe. Google announced its own Agent Payments Protocol in September 2025 with more than 60 launch partners including Mastercard, PayPal and American Express, per Google Cloud. Perplexity added agent-driven checkout in its Comet browser with PayPal, per Digital Commerce 360.
Is agentic commerce real momentum or hype?
It is real, and it is early. Grand View Research valued the agentic commerce market at $5.7 billion in 2025 and projects it reaching $65.5 billion by 2033, a compound growth rate of about 35.7%, per Grand View Research.
The hype part is the timeline. Most people are not yet handing their card to an agent for a considered purchase, and plenty of the season's AI influence was recommendations and support, not fully autonomous buying. But the direction is not in doubt. When Etsy, Shopify, Stripe, Google, PayPal and the card networks all build for the same behaviour in a single year, they are not guessing. They are paving a road they expect traffic on.
Does agentic commerce kill retention marketing?
No. It raises the stakes on retention, because the part of the journey an agent is best at taking from you is the top of it: discovery and the one-off sale. What an agent cannot take is the relationship you already have with someone who has bought from you and knows your brand.
Think about where an agent adds the most value for a shopper. It is the cold, comparison-heavy first purchase, the "find me the best X" moment. That is exactly the acquisition work brands sink most of their budget into, and it is now being commoditised by a layer that sits between you and the buyer. When acquisition gets commoditised, the margin and the defensibility move to whoever owns the ongoing relationship.
When an agent owns discovery, the customer you already have is the only audience nobody can bid you out of.
Why does keeping the customer matter more now?
Because the economics were always better on retention, and agentic commerce widens the gap. A 5% lift in customer retention can raise profit by 25% to 95%, the classic Bain finding still cited across the industry, per Ringly. Acquiring a new customer can cost anywhere from 5 to 25 times more than keeping one, per Harvard Business Review.
And repeat buyers carry the business. Repeat customers account for roughly 48% of ecommerce transactions and spend around 67% more on average than new ones, per Ringly. Yet the average repeat purchase rate sits around 28.2%, per Sender, which tells you most brands leave the second and third order on the table. That gap is the opportunity. As agents skim the first purchase, the second one is where you get to be the brand the customer chose, not the option an agent surfaced.
The blast
- One newsletter to everyone on Tuesday, whether the moment fits or not
- A refill nudge that lands two weeks after the jar ran out
- Win-back sent to people who never left, and skipped for the ones drifting
- Discovery and the first sale handed to whoever the agent picks
One buyer at a time
- Each customer read on their own timing, history and behaviour
- The refill lands the week it is actually due, in their preferred channel
- A quiet regular gets a genuine reason to come back before they slip
- The relationship after the first sale becomes the moat
What does a brand's own agentic retention layer look like?
It is an agentic system that works your existing customers one at a time and decides the next best move for each person, on your data and in your voice. Where a shopping agent optimises for the shopper across every brand, your retention layer optimises for your relationship with your customer, which is the one thing no external agent is trying to protect.
Take the coffee bag running thin. A customer who bought a 1kg bag six weeks ago is due a refill about now. A blast cannot see that. An agentic layer reads the signal per person, decides the timing is right, writes the message in your voice, and gets it checked before it goes. The marketer sets the goal in plain English and approves what goes out. The AI just removes the hours cap that stopped a small team from doing this for every customer.
Morning Sam, your Ethiopia bag is about to run thin
Going by when you last stocked up, you are near the end. Want us to send another before the weekend so you are not stuck with instant on Saturday? Same grind as last time.
Reorder the sameUnder the surface, four agents share the work for each customer, with the marketer in charge of the goal and the sign-off.
Won't the AI agents just handle retention too?
Not for you. A shopping agent works for the buyer and stays neutral across brands, so it has no reason to defend your margin, your voice, or your customer's loyalty to you specifically. Retention on your terms has to run on your own data, across your own channels, pointed at your own goals.
That means the brands that come out ahead are the ones that treat their owned relationship as infrastructure, not an afterthought. Email, SMS, push, WhatsApp, in-app, ads, and support all working from one read of the customer, so a person gets one coherent brand across every touch instead of a different message from every tool.
How would you know it is actually working?
You hold back a control group and measure against it. A brand sets aside a slice of customers who do not receive the agentic layer, and the lift is measured as the difference between the two groups, so the number you see is caused by the work and not by the season or a good month.
On category economics, that kind of per-customer engagement models out to as much as 50% more revenue. That is a modelled figure measured against the control group a brand sets, not a promise, and the honest way to read it is as the size of the prize if the relationship is worked properly, not a guarantee you can bank on day one. The price is tied to what that control group shows: PilotX takes 10% of the extra sales it adds, nothing if it adds nothing, capped at $2,500 a month.
Agentic commerce is going to keep pulling discovery into chat, and that is fine. The first purchase was never where the durable value lived. The value is in the customer who comes back, and that is a relationship you can still own, if you build the layer to work it. The quickest way to see what you are leaving on the table is to look at your own numbers.
Find the gaps in your retention with the free Revenue Leak Audit. It reads your funnel and shows where repeat revenue is slipping, in about the time it takes to finish a coffee. If you want to see the working over a fortnight, the 14-day recovery pilot is there too.