A signal worth watching
In its January 12, 2026 analysis of the 2025 holiday season, Adobe reported that traffic from generative AI tools to U.S. retail sites rose 693.4% year over year. Adobe describes those referrals as people clicking from AI tools to retail websites.
That distinction matters. A referral is not proof that an agent completed a purchase autonomously. A large growth rate also does not tell us what share of all customers used AI. It tells us that this particular route into retail websites grew quickly during the period Adobe measured.
Translate carefully
A neighborhood bakery, a plumber and a national retailer do not have identical customer journeys. Holiday retail data cannot tell an owner how many catering inquiries they will receive next month, or whether their customers want an assistant to handle the entire request.
Our interpretation is more practical: it is worth checking what happens when a customer arrives through an assistant. Can the visitor understand the offer? Are the relevant rules easy to find? Is the next step a usable form, a working ordering link or a clear contact route?
Run a small experiment
Pick one service and one question that customers already ask. Record the answer produced from current public information. Improve the underlying facts and request instructions, then repeat the test with the same question. Keep the original result so the comparison is inspectable.
If you also measure real traffic, distinguish referrals, started forms, completed inquiries and accepted work. Avoid counting all four as sales. The useful response to a growth headline is a proportionate experiment with a defined outcome. Small businesses can learn from a changing customer journey without buying into a prediction that every customer will become an agent user.
Sources & further reading
Sources checked 2026-10-12. Recommendations and examples reflect Welvia’s approach; they do not imply a partnership or guaranteed result.