RESEARCH & SOURCES
The findings behind the percentages.
myuser.ai’s analysis examines two external findings: 98% more gross profit per customer with AI daily newsletters versus no daily newsletter at Wine Access, and 21.7% higher conversion with AI chat versus an unavailable-support message at another retailer. Below, we explain those comparisons, separate the authors’ findings from our interpretation, and link to both original papers.
In this guide
Email newsletters: one Wine Access trial
Peer-reviewed paper, published January 13, 2026: Large Language Models and Creative Content Design: a case study of email marketing at Wine Access.
Randomized trial, January 15–28, 2024: 27,500 customers; 9,000 each received human, AI or hybrid newsletters; 500 controls missed daily newsletters. All groups received 30 other non-newsletter emails.
AI gross profit/customer: $10.222 versus $5.158 for controls. Our Table 3 calculation: (10.222 ÷ 5.158 − 1) × 100 = 98.18%, rounded to 98%.
Two outlier wines priced above $400 were excluded from the analysis. Humans fact-checked AI copy. This tested neither personalization nor cart reminders. AI did not significantly outperform human writers.
Customer replies: a presale-chat experiment
Preprint v6, dated June 30, 2026: Generative AI and Sales Productivity: Field Experiments in Online Retail.
One cross-border retailer randomized 44,614 presale-chat customers during September–October 2023. AI replies increased conversion by 21.7% relative, or 1.31 percentage points, compared with an automated message saying support was unavailable.
AI alone did not significantly outperform human replies. This was retail chat, not SMS or myuser.ai. Two authors consulted for the retailer; another was its employee. Preprint findings may change.
Source: Lu Fang, Zhe Yuan, Kaifu Zhang, Dante Donati & Miklos Sarvary (2026 preprint)
What myuser.ai takes from these findings
These experiments make the comparison worth examining. The Wine Access result measures daily newsletters versus no daily newsletters. The chat result measures available AI replies versus an unavailable-support message. Neither comparison establishes that AI beats a human.
For our demos, the useful question is whether a timely message or available answer helps a shopper take the next step. That is our interpretation of the research, rather than another measured finding. Establishing the lift from myuser.ai, personalized reminders or SMS requires an experiment with that service and its customers. Compare orders and costs against what would have happened without the message.