Visa's 43% AI Miss, Fanatics' 10% Lift & First Insight
Visa reveals a 43% AI shopping failure rate, Fanatics reports 10% higher engagement, and new AI tools help fashion brands make better product decisions.
– 2 min read
Visa reveals how often AI shopping agents get purchases wrong, Fanatics reports stronger engagement, and new tools bring customer intelligence and automated photography into fashion. Everything worth knowing in AI Commerce in two minutes:
01 / Visa finds AI shopping fails 43% of the time.
On November 5, Visa published a benchmark of 22 AI models across 15 shopping challenges. Agents failed 42.8% of test runs, and 54% of those failures still ended in a purchase.
Insight: AI recommendations still need stronger product understanding.
02 / First Insight brings customer intelligence into AI.
On November 5, First Insight announced tools connecting real shopper feedback, purchase intent and predicted pricing directly to AI assistants. Fashion teams can use the insights before committing to production.
Insight: Better product decisions start before inventory exists.
Read First Insight’s announcement →
03 / Fanatics reports 10% higher engagement with AI.
Fanatics launched an OpenAI-powered version of Shop the Look on October 26. Testing across approximately 30,000 products delivered 10% higher engagement than its previous recommendation experience.
Insight: Smarter merchandising can increase engagement without more traffic.
04 / Koozee automates fashion content production.
On November 5, Koozee introduced reusable AI workflows for apparel photography and video. Fashion teams can generate product imagery, variations and campaign assets across entire collections.
Insight: Creative production is becoming scalable without traditional photoshoots.
SPOTLIGHT
AI can design your next collection. But will anyone buy it?
Fashion brands can generate hundreds of product concepts in hours.
Deciding which deserve production is still much harder.
First Insight’s latest announcement brings consumer testing into the AI tools merchants already use. Instead of relying entirely on historical sales, teams can access shopper feedback and modeled willingness to pay before placing orders.
That’s an important distinction.
An AI-generated design isn’t evidence that customers want it. And a prediction isn’t a guarantee of sales.
The opportunity is combining faster product creation with better validation.
Imagine testing ten potential designs before committing to samples, production and marketing.
The commercial impact should be measured in full-price sell-through, markdown exposure and gross margin.
Creating more products is easy. Choosing the right ones creates value.
TIP FOR TODAY
Take five minutes to challenge one product decision.
- Choose an upcoming product from your next collection.
- Ask what evidence suggests customers will actually buy it.
- Separate historical sales data from current customer feedback.
Find one assumption worth validating before production.
DEAISY | Your Daily AI Advantage in Commerce.
Written by
Co-Founder and COO.
Agent at OpenAI.
Comments
