Every year, Q4 feels like e-commerce’s final exam. But this year, the test has a new examiner: AI – artificial intelligence.
We are entering a peak season where record online sales coexist with slowing growth, rising consumer debt, margin pressure, and an AI arms race that very few sellers are genuinely prepared for.
I want to write about this moment not to celebrate another “record Cyber Week” headline, but to ask whether the system underneath those numbers is healthy, fair, and future-proof for the broader e-commerce community.
The Numbers Look Strong, but the Story is Fragile
Let’s start with the data.
In the 2024 holiday season, U.S. consumers spent $241.4 billion online between 1 November and 31 December, an 8.7% year-on-year increase and a new record. More than half of those transactions (54.5%) happened on smartphones. Cyber Monday alone delivered $13.3 billion in online sales, up 7.3% from 2023, with shoppers spending nearly $15.8 million per minute at the peak. Black Friday reached $10.8 billion, up 10.2% year-on-year.
For 2025, Adobe now projects U.S. holiday online sales of about $253.4 billion, a more modest 5.3% increase – still growth, but clearly slower than last year’s 8.7%. Digital Commerce 360, aggregating forecasts from Adobe, Deloitte and Salesforce, suggests that 56.1% of online holiday sales this year will come from mobile and that Buy Now, Pay Later (BNPL) spending could reach $20.2 billion, up from roughly $18.2 billion last season.
In other words, the Q4 engine is still running, but on deeper discounting, more credit, and more mobile-driven impulse. For marketplaces, platforms, and payment providers, this is good news. For smaller merchants, brands, and the logistics ecosystem, the picture is much tougher. Higher operational costs, tighter margins, fierce price competition, and increasingly “trained” consumers who expect 25–30% discounts as a baseline, not a special offer.
November Deals: No Longer Events but an Operating System
November used to be defined by a handful of big days: Singles’ Day, Black Friday, Cyber Monday. That calendar is now effectively a continuous promotion window.
In China, this year’s Singles’ Day (Double 11) campaign again delivered record sales for Alibaba and JD.com, despite a sluggish macro environment, driven heavily by AI-driven promotions, targeted marketing, and chatbots. More than 34,000 brands doubled their sales year-on-year on Alibaba’s platforms, and thousands of brands recorded triple-digit growth thanks to algorithmic optimisation.
In the U.S., Adobe, Salesforce and others all highlight the same structural shift:
- Discounts start earlier, often in October.
- Shoppers increasingly wait for the “right” price, compressing margins.
- BNPL usage hits records, adding volume but also risk. BNPL drove $18.2 billion in online spend in 2024 (+9.6% YoY), and is forecast to grow again this year.
At the same time, returns are exploding. Salesforce reports that global online sales reached $1.2 trillion in 2024, with $282 billion in the U.S.. Still, returns now stand at 28% of orders, up from 20% a year earlier, a significant drag on profitability for sellers and brands.
So when we celebrate “record November deals,” we should add an asterisk; those numbers hide mounting structural stress for everyone who actually has to hold inventory, ship orders, and absorb returns.
The AI Q4: From Hype to Hard Power
Now to the real game-changer: AI.
A year or two ago, “AI in e-commerce” was mainly a slide in a pitch deck. In Q4 2025, it has become infrastructure:
- One in three U.S. consumers (33%) now say they plan to use generative AI in their holiday shopping journey, more than double the 2024 figure. They use it to find the best deals, summarise product reviews, and generate shopping lists.
- Salesforce’s latest projections suggest that AI and AI agents could influence up to $263 billion in global online sales in 2025, including around $51 billion in the U.S.
- McKinsey estimates that generative AI could unlock $240–390 billion in annual value for retailers, translating into a 1.2–1.9 percentage-point margin uplift if implemented at scale.
AI is no longer an abstract future. It’s already visible in:
AI search and merchandising: Where product ranking is not just about keywords but multilayered behaviour signals, vector search, and real-time intent scoring.
AI agents and chatbots: Which steer customers to higher-margin products, nudge basket size, and automate service at scale. Salesforce data from the 2024 season already showed that AI-influenced purchases represented $229 billion globally, with chatbot usage up more than 40% year-on-year.
Content at scale: Product descriptions, creative variations, personalised emails, and ad copy are now being generated or heavily assisted by AI, compressing production cycles from weeks to hours.
This is why I believe Q4 is becoming the real dividing line between AI-enabled and AI-exposed sellers.
If you are a marketplace, a significant platform, or a large retailer, you are already experimenting with AI for forecasting, personalisation, inventory allocation, and media buying. Suppose you are a small or mid-size brand. In that case, you are increasingly playing inside someone else’s algorithmic universe, where AI systems mediate visibility, pricing, and even customer relationships you do not control.
Who Wins and Who Gets Squeezed?
Let’s be direct: AI is currently structured to concentrate power, not distribute it.
Winners:
- Big platforms and marketplaces that own first-party data at scale and can afford their own models or deep integrations with cloud AI providers.
- Brands with strong data discipline (clean product feeds, clear attribution, robust experimentation) that can actually benefit from algorithmic optimisation rather than be randomly boosted or buried.
- Logistics and fulfilment operators that use AI to optimise routes, staffing, and inventory, reducing last-mile costs and shipment delays during peak windows.
Those at risk:
- Small sellers who treat AI as just another “tool” rather than a strategic shift. They might use a bit of generative copywriting or image editing, but their product catalogue, pricing, and customer data remain fragmented and under-leveraged.
- Merchants are overly dependent on BNPL-fuelled demand, where any regulatory shift or macro shock could rapidly reduce conversion while leaving them with the cost of fraud and returns. Evidence is already emerging that BNPL can push vulnerable consumers into debt; one recent analysis found that roughly 24% of BNPL users missed at least one payment in 2024, amid mounting fees and financial stress.
- Communities whose participation is reduced to “content and clicks” but who do not have ownership of the data or the margins. Influencers and social-commerce sellers generate demand, but capture only a thin slice of the value chain while shouldering reputational risk and volatility.
We sit at the intersection of sellers, enablers, and policymakers. Our community spans marketplaces, logistics providers, fintechs, agencies, and thousands of merchants trying to grow across borders.
We are all watching the same Q4 trends:
- Record online sales but slower growth.
- Heavier reliance on BNPL and promotions to keep volumes up.
- Higher return rates are eating margins.
- AI is moving from “nice-to-have” to core infrastructure for search, pricing, and customer engagement.
If we treat this purely as a growth story, we are missing the point.
I am writing this op-ed because I believe Q4 2025 is the season in which global e-commerce must choose a direction, and these questions need to be carefully answered.
- Will AI be used mainly to extract more value from the same constrained consumers, via hyper-targeted promotions, frictionless debt, and infinite content, or to build healthier, more predictable, and more inclusive ecosystems for sellers and buyers?
- Will marketplaces continue to absorb almost all of the AI productivity gains in their own margins, or will some of that efficiency be shared with merchants in the form of better tools, lower fees, and more transparent algorithms?
- Will policymakers remain reactive, stepping in only after debt crises, unfair practices, or data breaches, or will they help set guardrails for BNPL, dark patterns, and abusive personalisation before we hit a breaking point?
What the Ecosystem Needs to be Arguing?
For the e-commerce community including platforms, logistics companies, payments providers, regulators, and trade bodies, Q4 2025 should not be about celebrating records alone. It should be about debating:
- Standards for transparent AI in e-commerce: Especially around search ranking, recommendation bias, and dynamic pricing.
- Responsible BNPL frameworks: Protect vulnerable consumers without killing innovation in payments.
- Cross-border data and tax regimes: Allow SMEs from emerging markets to actually compete in AI-driven marketplaces instead of being permanently locked behind algorithmic and compliance walls.
These debates must not be postponed to after the season. They are already shaping who wins this Q4 and who will still be standing when the return season is over.
Q4 used to be about who shouted the loudest and discounted the deepest. In the age of AI, it is increasingly about who owns the data, the models, and the levers of personalisation.
If we want an e-commerce ecosystem that is not only bigger, but also fairer and more resilient, this is the moment to say so clearly, with numbers on the table, and without being blinded by record sales headlines.
And that is precisely why I am writing this now. Q4 – Q4 – Q4 – Q4 – Q4 – Q4 – Q4
BURAK YALIM
Editor in Chief – WORLDEF E-Commerce Magazine
