Retail and e-commerce: the full guide

A retailer makes money on the gap between what it pays for goods and what shoppers pay, minus everything it costs to hold, move and sell them. Most of the cost is committed before the shopper decides: the goods are bought, the stores are rented and the staff are scheduled.

Five pools of money decide the year.

Online sales were 17.1% of US retail sales in the second quarter of 2026. The rest went through stores and other channels. Marketplaces are a channel and a competitor at once. On Amazon, more than 60% of sales come from independent sellers, as reported by Amazon.

AI touches margin through pricing and markdowns, inventory through forecasting, cost through store labor and fulfillment, returns and shrink through fraud screening and cameras, and channel mix through the AI assistants that decide which products shoppers see.

Gross margin: sharper prices, with new rules. AI pricing tools adjust prices to demand, stock and competitors. The new state laws separate prices set from market conditions, which are generally not their target, from prices set with a shopper's personal data, which several states now restrict or require to be labeled.

Inventory: better forecasts, if someone acts on them. McKinsey reported in 2021 that early adopters of AI in supply-chain management had 35% better inventory levels and 65% better service levels than slower competitors. That is a consulting firm's comparison, not a controlled test. Run your own.

Returns: fraud screening. Retailers estimated that 9% of returns are fraudulent, and 85% of large online merchants surveyed said they use AI to fight it. The risk is flagging good customers as cheats.

Stores and loss prevention: cameras that can misfire. The FTC found that a drugstore chain's facial recognition produced false matches that led staff to accuse innocent shoppers, and that the chain never tested its accuracy.

Product pages and reviews: cheaper, with a legal line. AI can draft product descriptions and answer shoppers' questions in minutes. A claim AI writes on your page is still your claim. The FTC's rule on reviews bans fake reviews, including AI-generated ones, and lets the agency seek civil penalties.

Discovery: AI assistants. Traffic to retail sites from AI assistants grew 138% in the year to May 2026, as reported by Adobe. The course Customers who shop through AI agents covers this channel in full.

Store staff: fewer checkout jobs. The Bureau of Labor Statistics projects cashier jobs to fall 6% from 2025 to 2035, citing self-checkout and online sales.

Time saved is not money saved. Faster work pays only if it lowers markdowns, stockouts, fraud or payroll.

Two patterns stand out. Most of the reported gains come from surveys run by sellers of AI or from the retailers themselves, so a retailer should ask for its own before-and-after numbers. And the regulators have acted where shoppers can be hurt directly: cameras that misidentify people and prices set with personal data.

Scenario 1: AI assistants complete a meaningful share of purchases in a category. This is a scenario, not a forecast. Shoppers ask an assistant and buy inside it, without visiting the retailer's site. Assumes: checkout inside assistants spreads beyond single items and a few sellers. Signal to watch: the share of your own orders that start in an AI assistant, and which assistants offer checkout in your category. Last checked: October 2026. OpenAI's checkout inside ChatGPT started with single-item purchases from US Etsy sellers, as reported by OpenAI. AI-referred traffic to retail sites grew 138% in the year to May 2026, as reported by Adobe.

Scenario 2: More states restrict personalized or algorithmic pricing. This is a scenario, not a forecast. Assumes: New York's broader ban becomes law and other states copy Maryland, Connecticut and New Jersey. Signal to watch: the New York governor's decision on the One Fair Price Act, the FTC's final policy statement, and New Jersey's pause on new electronic shelf labels from February 1, 2027. Last checked: October 2026. New York's legislature passed the ban in June 2026, and it was awaiting the governor's signature as of October 10, 2026.

Scenario 3: Return fraud and organized theft push more stores toward AI cameras. This is a scenario, not a forecast. Assumes: retailers keep reporting theft and fraud pressure even as police reports fall. Signal to watch: retailers' own shrink numbers, state rules on biometric data, and FTC action on surveillance tools. Last checked: October 2026. Retailers estimated 9% of returns were fraudulent in 2025. Police-reported shoplifting fell 10% in 2025 in 21 cities, and the researchers say police data undercount it. The FTC's five-year ban on one chain's facial recognition shows what follows untested cameras.

All three reward the same preparation: clean product and stock data, a written rule on what may set a price, and a human check before any camera alert leads to action.

Gain: retailers with clean product and stock data. AI assistants and forecasting tools can only use what they can read.

Gain: large chains and marketplaces, for now. In one study, AI assistants led with a large chain about 2.5 times as often as with a small store.

Gain: shoppers, when forecasts work. Fewer empty shelves and fewer clearance piles help shoppers and retailers alike.

Lose: retailers whose prices look personal. The FTC says shoppers expect the listed price to be the price everyone sees, and that undisclosed personalized prices may break the law.

Lose: stores that act on untested cameras. One chain lost the use of facial recognition for five years.

Mixed: store staff. Cashier jobs are projected to fall 6% from 2025 to 2035, while stores still need people for service, stocking and online orders.

Two points still contested.

1. Does AI pricing help shoppers or harm them? The case that it helps: a study of 114 grocery stores found electronic shelf labels did not bring surge pricing. Short-lived price jumps touched about one product in 20,000 before the labels and rose by only 0.0006 percentage points after. Prices mostly followed wholesale costs, competitors and manufacturer promotions. The case that it harms: the FTC's chairman said in August 2026 that shoppers expect the price they see to be the price everyone sees, not an estimate of what they will pay based on their data. The FTC noted that informed shoppers might avoid retailers that personalize prices. The two sides are about different things. The study covers shelf prices that change for everyone. The FTC's concern is prices set for one person.

2. Will AI assistants favor big retailers or give small sellers a fair shot? The case for big retailers: in a study commissioned by Lightspeed, which sells retail software, large chains took 46% to 58% of top recommendations when assistants searched the web. The case for small sellers: OpenAI says ChatGPT's product results are unsponsored and ranked on relevance, and its checkout launched with US Etsy sellers. The same study found small and large stores each made up about 38% of cited sources, and adding the word "independent" to a request raised the share of small shops named from about one-third to nearly four-fifths.

The cost ranges are illustrative and will be replaced with reviewed figures. The order matters more than the amounts. The first three rows save money where a mistake is fixed quietly. The last two touch shoppers directly and carry legal risk.

One test before any contract. Ask the vendor to run its tool on last season's data for a set of your stores and compare its answer with what actually sold.

Suppliers. Consumer goods makers, distributors and beverage producers use AI to forecast demand and plan promotions, and they ask retailers for more sales data to feed it. See the guides for consumer goods, wholesale distribution, and beer, wine and spirits.

Marketplaces. More than 60% of sales in Amazon's store come from independent sellers, as reported by Amazon. A marketplace's own ranking rules decide which products shoppers see first.

AI assistants. Assistants now sit between shoppers and stores, and in one study they led with large chains more often than small ones. See the course Customers who shop through AI agents.

Regulators. State attorneys general enforce the new pricing laws, and the FTC enforces rules on deception, reviews and surveillance. Pricing and camera vendors hold your shoppers' data, so your contracts must say what data they may use.

October 13, 2026: first published.

October 2026 (reflected at launch): New York's ban on personal-data pricing, passed by the legislature in June 2026, was still awaiting the governor's signature as of October 10, 2026. Affects sections 2, 4 and the decision exercise.

September 2026 (reflected at launch): a study commissioned by Lightspeed found AI assistants led with large chains in 46% to 58% of answers when searching the web. Affects sections 3 and 5.

August 2026 (reflected at launch): the FTC proposed a policy statement treating undisclosed personalized pricing as possible deception. Affects sections 2, 5 and the decision exercise, and sets the guide's main conclusion: write the pricing rule before buying the pricing tool.

July 2026 (reflected at launch): New Jersey enacted a ban on personal-data pricing for groceries and a one-year pause on new electronic shelf labels from February 1, 2027. Affects sections 4 and the decision exercise.

June 2026 (reflected at launch): AI-referred traffic to US retail sites grew 138% in the year to May 2026, as reported by Adobe. Affects sections 1, 2 and 4.