Roughly 8 in 10 merchants now use at least one AI feature in their store, according to the Shopify Commerce Trends 2025 report. That figure is the headline a year of AI-in-commerce coverage has been circling. The interesting part is what those features actually do in practice. The eCommerce team using AI is not running a mysterious autonomous system. They are running four or five well-understood workflows: product copy, image generation, on-site search, customer support, and inventory forecasting. The shift in the role is in which of those workflows now run on a draft-and-edit cycle instead of a write-from-scratch one.
The four workflows that pulled most of the adoption
Shopify’s report and the merchant-facing telemetry from third-party app stores point at the same handful of use cases doing most of the work.
Product descriptions are the front door. A team that used to handwrite each PDP now feeds the supplier sheet into ChatGPT or Claude with a brand-voice example, gets a draft, edits, and ships. The first weeks of doing this are the steepest learning curve in the role, in that it forces a merchandiser to articulate what their brand voice actually is in writing, and most teams discover their existing brand voice doc is too vague to drive a model.
Product photography and lifestyle imagery is the second front. Tools like Midjourney, Gemini, and Shopify’s own Magic suite are now standard for category banners, ad-creative variants, and lifestyle backgrounds for product shots. The work that used to require a half-day photo session and a week of retouching now takes a morning. The gating factor is brand judgement on whether the AI-generated images feel native to the store or feel like stock.
Search and recommendations is the third. AI-powered semantic search (where a customer types “warm jumper for layering” and gets results that do not contain the word “warm”) has materially changed conversion patterns on stores that have it. This is mostly invisible work for the team, in that it is a switch the platform offers and a tuning loop afterwards.
Support is the fourth. The cross-sector data on AI customer service is striking: roughly 65% of incoming queries are now resolved without human intervention in 2025, up from about 52% in 2023, according to industry trackers. For eCommerce specifically that means delivery questions, returns initiation, and order lookups are now overwhelmingly self-serve. Human time moves to the harder cases.
Where the savings go, and where they don’t
The saved hours rarely come out of headcount on a small store. They go back into the work that grew the team in the first place: more SKUs launched per week, more campaign variants tested, more category pages refreshed seasonally instead of yearly. On larger merchants the picture is more mixed. Bigger teams have started consolidating roles where copywriting, basic image production, and tier-one support overlap with AI-handled work.
The job category that has actually contracted is the freelance market for entry-level product-description writing. The mid- and senior-level eCommerce roles, where the work is editorial judgement and merchandising strategy, are growing roughly in line with overall ecommerce employment.
The productivity gain is real but variable
The “AI saves 8 to 10 hours a week” figures floating around eCommerce marketing content are not from controlled studies. The closest evidence base is the GitHub Copilot RCT showing a 55.8% speedup on a benchmark coding task, and the Harvard Business School and BCG consultant study showing roughly 25% reduction in time-to-completion on knowledge-work tasks within scope. Together those bracket the realistic productivity gain on eCommerce-specific work in the 20 to 40% range on the tasks AI is actually good at.
That still leaves the majority of an eCommerce team’s week on tasks AI is not yet good at: negotiating with suppliers, working through a returns dispute that has gone sideways, reviewing the analytics on a campaign that underperformed and figuring out why, deciding which markets to expand into next.
What the next 12 months look like
The Shopify trend data points at three movements through 2026. First, the share of merchants using AI features will keep climbing past the current 8 in 10, with the long tail being the smallest stores and the most heritage-conservative brands. Second, the gap between merchants who use AI for one feature and merchants who have integrated AI across copy, imagery, search, and support will widen. The 80% adoption headline hides a long-tail distribution where the top quartile uses many features and the bottom uses one.
Third, the most interesting product launches in 2026 will be agentic, with merchant-facing tools that can take a brief and output a full campaign (copy, image, audience targeting, scheduling) rather than just one of those pieces. The McKinsey State of AI 2025 report shows 23% of organisations already scaling agentic AI somewhere in their operations. eCommerce is one of the better-fit use cases because the workflows are well-defined and the outputs are easy to evaluate quickly.
For people working in eCommerce roles, the practical question is no longer whether to learn AI tooling. It is whether your team is moving from “we use ChatGPT for product copy” to “we run a tested, branded, repeatable workflow that any new hire can pick up.” The merchants getting genuine commercial uplift from AI are doing the second. The ones who say they are using AI but cannot describe their workflow are still doing the first.