# Evidence: Commerce 7 of 108 entries in the collection “Evidence” by Robert Haase, as of 21 September 2026. Page: https://robert-haase.de/en/evidence.html · Overview of all claims: https://robert-haase.de/en/evidence.md · JSON: https://robert-haase.de/en/evidence.json · Deutsch: https://robert-haase.de/belege-handel.md License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Please cite the primary source, not this page. This file is generated from the page. Where the two differ, the page applies. ## How this collection is built Every figure is traced back to the body that measured it, not to the article citing it. On their way through the retellings, figures lose their denominator first, then their caveat, and finally their origin. Where a figure is only accessible through a third party, that intermediary is named in the source line. Own measurements carry their method with them; they have not been independently verified yet. **The limit belongs to the number.** The most common error is not the wrong number but the right one carrying a claim that reaches further than the evidence. That is why every entry has two parts, and the second one matters more. Above each figure sits what kind of evidence it is, from verified study to single case. That decides how far it carries. What does not survive the check does not get in, or gets taken out, my own articles included. One of them claimed that 44 percent of US online shoppers begin their purchase journey in a language model, attributed to Bain. Bain gives two other figures, 17 percent and 30 to 45 percent, which had merged into one along the way. Both are here now; the 44 is not. This page ages. Every entry carries its date; superseded numbers get replaced, not quietly deleted. If you find an error, [write to me](mailto:hallo@robert-haase.de) and I will correct it and note the date. The collection does not map the state of the research, only the figures I needed for my own texts. Free to use with attribution. When in doubt, link the primary source rather than this page. ## Grades in this topic + Verified study, vendor documentation, or court decision (2) → airline-direktkanal, shopify-knowledge-base + Preliminary: prototype, single test, forecast, or vendor figure (0) + Status, case report, or market observation (5) → checkout-rueckbau, walmart-verhandlung, journey-start, kaufentscheidung, ucp-gremium-ohne-zahlen --- ## checkout-rueckbau **Claim:** Buying directly inside the chat was rolled back a good five months after launch. Live at that point were either a dozen or close to thirty Shopify merchants, depending on the source. **What follows and what does not:** The thesis that commerce moves wholesale into the chat is refuted for now. What is not refuted is the shift in discovery: found at the agent, bought at the brand. **On the number:** the source gives two conflicting figures, a dozen per a press report and "closer to 30 and climbing" per Shopify itself. Both cover Shopify merchants only, not all partners. **Source:** Forrester, analysis of the instant checkout rollback · 7 March 2026 (launch: 29 September 2025) · [Source](https://www.forrester.com/blogs/what-it-means-that-the-leader-in-agentic-commerce-just-pulled-back/) **Grade:** Market observation · Group: Status, case report, or market observation **Permalink:** https://robert-haase.de/en/evidence.html#checkout-rueckbau --- ## airline-direktkanal **Claim:** In a flight-search test, language models went to the airline’s own website directly in only about five percent of cases. They preferred booking portals. **Why this matters for brands:** The direct channel and the loyalty program, the most expensive distribution assets many brands own, are bypassed in the agent layer. The study’s explanation is remarkably unglamorous: portals deliver cleaner, more structured, more agent-readable data. **Limits:** three European carriers, their ten most relevant markets, three language models, 60 non-branded prompts. One industry, one point in time, no claim about other categories. **Source:** Bain & Company, test across three carriers, three language models, and 60 prompts · 16 March 2026 · [Source](https://www.bain.com/insights/is-the-airline-industry-ready-for-agent-led-bookings/) **Grade:** Controlled test · Group: Verified study, vendor documentation, or court decision **Permalink:** https://robert-haase.de/en/evidence.html#airline-direktkanal --- ## walmart-verhandlung **Claim:** Walmart has supplier negotiations run by an AI system: 2,000 negotiations at once, around three percent average savings, and three out of four suppliers preferring to negotiate with the machine over a person. **Where the numbers come from:** the system’s vendor and Walmart itself, reported via Bloomberg. There is no independent audit. The closing rate of 68 percent often quoted alongside comes from an earlier Harvard Business Review case study. **The third number remains the striking one:** the preference for the machine comes from the other side of the table, not from the operator. That is mandate in procurement, quantified and in production. **Source:** Pactum figures via Bloomberg, April 2023 · closing rate: Harvard Business Review, November 2022 · [Source](https://www.engadget.com/walmarts-suppliers-would-rather-negotiate-with-ai-than-a-human-162131831.html) **Grade:** Vendor and company figures · Group: Status, case report, or market observation **Permalink:** https://robert-haase.de/en/evidence.html#walmart-verhandlung --- ## journey-start **Claim:** 30 to 45 percent of US consumers use generative AI for product research and comparison. 17 percent said they would begin their holiday shopping on an AI platform. **What the numbers do not say:** They rest on self-reporting and cover the US, not the German-speaking market. **And the two figures measure different things:** research and comparison is not the same as beginning the shopping journey. Merging them produces a number that does not exist. A journey begun is also not a purchase. As an order of magnitude for the shift in discovery they are usable; as a revenue statement they are not. **Source:** Bain & Company, Consumer Lab survey, US · November 2025 · the May 2026 insights page states around 30 percent · [Source](https://www.bain.com/about/media-center/press-releases/20252/agentic-ai-poised-to-disrupt-retail-even-with-50-of-consumers-cautious-of-fully-autonomous-purchasesbain--company/) **Grade:** Survey · Group: Status, case report, or market observation **Permalink:** https://robert-haase.de/en/evidence.html#journey-start --- ## kaufentscheidung **Claim:** Only 11 percent of respondents would let an AI make the purchase decision, and only in low-stakes categories such as personal care and household supplies. **What puts the number in perspective:** Even willingness to accept mere *help* is not a majority: 31 percent would let AI narrow choices for household supplies, 28 percent for electronics. It is the counterweight to forecasts about autonomously buying agents. Stated preference and later behaviour do diverge regularly, especially where experience is missing. Sample: 322 US consumers. **Source:** Gartner, survey of 322 US consumers in January 2026 · published 27 May 2026 · [Source](https://www.gartner.com/en/newsroom/press-releases/2026-05-27-gartner-survey-finds-consumers-want-ai-shopping-help-but-not-ai-purchase-decisions) **Grade:** Survey · Group: Status, case report, or market observation **Permalink:** https://robert-haase.de/en/evidence.html#kaufentscheidung --- ## ucp-gremium-ohne-zahlen **Claim:** The Shopping Tech Council of the UCP commerce protocol has 16 seats; since 24 April 2026 they include Amazon, Meta, Microsoft, Stripe and Salesforce alongside Google, Shopify, Etsy, Target and Wayfair. How many merchants actually run the protocol is stated by none of the companies involved. Google names example merchants — Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair, and Shopify merchants such as Fenty and Steve Madden — attached to the word “soon”. **What the number does not say:** A seat on a council is not an implementation. That Amazon, Meta and Microsoft help shape the protocol says nothing about whether they use it in their own stores. **The missing adoption figure is a negative finding:** it is absent from the primary sources checked — the UCP announcements in the project repository and two Google posts from 19 March and 19 May 2026. It may exist elsewhere. **A third-party count exists:** the checking service UCP Checker reports 15,735 verified storefronts among 19,336 monitored domains on 26 August 2026; it counts what stores declare technically and what responds, not orders. Besides the Shopping Tech Council there are now separate councils for food ordering, lodging and payments (as of 21 September 2026). **Care with secondary sources:** they widely reproduce Google's sentence without the “soon”, turning an announcement into a fact. **Source:** Universal Commerce Protocol, project repository, announcement of new Tech Council members, 24 April 2026 · Google, “Universal Cart”, 19 May 2026 and “UCP updates”, 19 March 2026 — neither with an adoption figure · third-party count: UCP Checker, “The State of Agentic Commerce — August 2026”, 26 August 2026 ([to the count](https://ucpchecker.com/blog/state-of-agentic-commerce-august-2026)) · councils per MAINTAINERS.md and GOVERNANCE.md in the Universal-Commerce-Protocol/.github repository, retrieved 21 September 2026 · [to Google’s announcement](https://blog.google/products-and-platforms/products/shopping/google-shopping-cart/) · [Source](https://github.com/Universal-Commerce-Protocol/ucp/discussions/379) **Grade:** Standards status · Group: Status, case report, or market observation **Permalink:** https://robert-haase.de/en/evidence.html#ucp-gremium-ohne-zahlen --- ## shopify-knowledge-base **Claim:** Since 16 May 2025, Shopify has offered merchants a free app of its own for deciding what AI shopping agents answer about their store. Merchants see automatically generated facts and common customer questions and can adjust answers or write new ones. The answers do not appear in the store; they serve AI platforms as a data source, and the app shows how many questions come from agents and whether the AI can answer them. **What this entry does not establish:** use or effect. These are the vendor’s statements without usage figures; the App Store listing shows 29 reviews averaging 3.6 of 5 stars. **What the app promises and what it does not:** According to Shopify it improves the accuracy of answers about the store, not how often the store appears in AI answers. The help page does not name the platforms that take up the answers, and the app is available in English only. **Source:** Shopify, help page “Shopify Knowledge Base” and the app’s App Store listing: launched 16 May 2025, free, developer Shopify, 29 reviews, language English · retrieved 21 September 2026 · [to the App Store listing](https://apps.shopify.com/shopify-knowledge-base) · [Source](https://help.shopify.com/en/manual/promoting-marketing/knowledge-base) **Grade:** Vendor documentation · Group: Verified study, vendor documentation, or court decision **Permalink:** https://robert-haase.de/en/evidence.html#shopify-knowledge-base --- End of file: 7 of 7 entries on Commerce. Last entry: shopify-knowledge-base.