AI Chat for Online Stores: Handle Sizing, Quotes, and FAQs Without Losing the Sale

Read a week of your store's chats and emails and you will notice something: most of them are the same twenty questions.

Will this fit me? Does it come in a larger size? When will it arrive? Can I return it? What would a bulk order cost? Each one is easy to answer, but each one takes a person a few minutes, and the shopper who doesn't get an answer quickly often leaves, or orders two sizes and returns one.

An AI assistant can answer most of these instantly. But a store assistant only works if it is grounded in your own data, knows when to hand off, and is introduced after you have fixed the pages that cause the questions in the first place.

Step Zero: Sort Your Last 500 Questions

Export recent chats, emails, and contact forms, and tag each one. You will usually find four groups:

Each group needs a different fix.

Fix the Page Before You Add the Bot

Some questions exist only because the page is confusing. A published case study about a beauty brand reports a direct signal of confusion: after a storefront refresh, the firm cites "a 61% reduction in U-turns and rage clicks on the homepage," meaning fewer visitors backtracking or clicking repeatedly in frustration. Every question a page answers clearly is one your assistant never has to.

Sizing: Where AI Pays for Itself

Fit questions are the most valuable to automate, because uncertainty turns into returns. A DTC wetsuit brand described the pattern: customers "hesitated at checkout without a smart sizing tool," and "often ordered multiple sizes and returned what didn't fit." Its developers added AI size advice to product pages, based on the brand's size finder measurements, alongside checkout fixes and conversion work. Completed checkouts rose 21% across all markets after those combined changes, and the firm reports a "significant reduction in sizing-related returns" without giving a figure.

Good sizing advice asks before it answers: height, weight, usual size in another brand, preferred fit. It should recommend from your real measurements, not from general rules.

Quotes: Turn Days Into a Conversation

Bulk and B2B requests often sit in email while someone works out a price in a spreadsheet. One DTC bag brand with a business recycling program said every recycling inquiry needed "manual processing through email chains and spreadsheet calculations." The assistant now in development asks for material, quantity, and location, then calculates the quote from the brand's rules. The same assistant serves retail shoppers and switches its tone based on who is asking.

The price itself should come from your pricing rules, never from the model's judgment. The model's job is to collect the inputs in a friendly way.

Guardrails Every Store Assistant Needs

What to Measure

Mistakes to Avoid

By the Numbers

Published industry case studies from software firms show where the time goes. The figures below are self-reported by the firms that published them, and each describes a single project rather than an industry average.

Start With a Systems Review

If your team answers the same questions all day, the first step is knowing which ones your data can answer for you. A systems review sorts your real customer questions, maps the data behind each answer, and ranks what to automate first. Explore our AI automation services or read more on the Delta blog.