Retailers expected 15.8% of 2025 sales to come back as returns, and shoppers now arrive through AI assistants.

A retailer buys goods, holds them, and sells them for more than they cost, through its stores, its own website and other companies' marketplaces. What it keeps depends on how much it marks down, how much comes back, how much goes missing, and what it costs to get goods to the shopper. Small changes in each of those decide the year.

Returns are a large leak. Retailers estimated that 15.8% of 2025 sales would come back, about $849.9 billion. For online sales the estimate was 19.3%. Among large online merchants surveyed, 85% said they use AI to detect or prevent return fraud.

Theft is real but hard to measure. Retailers reported shrink, meaning goods lost to theft, error and unknown causes, of 1.6% of sales in fiscal 2022. The industry's best-known dollar estimate of organized theft was withdrawn in 2023 after a trade publication found it rested on old data. Police-reported shoplifting fell 10% in 2025 across 21 large cities, though police data undercount it.

Shoppers are arriving through AI. Traffic to US retail sites from AI assistants grew 138% in the year to May 2026, and those visits converted 54% better than other traffic, as reported by Adobe. Online sales were 17.1% of US retail sales in the second quarter of 2026.

Retailers say AI is paying. In a survey by NVIDIA, which sells AI chips, 89% of retail and consumer goods respondents said AI raised revenue and 95% said it cut costs. A seller's survey of its own market is a weak test. Your own numbers are the strong one.

Pricing is where the rules are moving fastest. New York requires a label on prices set with a shopper's personal data. Maryland, Connecticut and New Jersey have passed limits on personal-data pricing in parts of retail, and New Jersey pauses new electronic shelf labels for a year from February 2027. The FTC proposed in August 2026 to treat undisclosed personalized prices as possible deception.

Cameras carry their own risk. The FTC banned a drugstore chain from using facial recognition for five years after false matches led staff to accuse shoppers of theft.

AI assistants favor big chains, for now. In a study commissioned by a retail software company, large chains took 46% to 58% of top recommendations when the assistant searched the web.

Time saved is not money saved. A better forecast pays only if it means fewer markdowns, fewer empty shelves or less stock to finance.

Here is what that looks like in practice. A retailer with $600 million in sales and a 35% gross margin earns $210 million in gross profit. One point of margin from sharper pricing is worth $6 million a year. If shoppers come to believe prices are set against them and 2% of sales leave, that is $12 million of sales and about $4.2 million of gross profit. The gain is real, and the loss can arrive faster.

What should a retailer's leaders do first? Three plain things.

1. Use AI first behind the counter: forecasting, replenishment, returns screening, and product content that a person checks.
2. Write a pricing rule before signing with any pricing vendor: what data may set a price, and when shoppers are told.
3. Check every camera, data and pricing tool against the rules in each state where you sell.

None of this needs a new team. It needs one leader who owns each answer.

The full guide shows where AI touches the money, what has happened, who gains and who loses, and what to ask before you spend.