Agent Usability Rating
Agent Walk
The passive scan tells you what an agent can see. The Agent Walk shows you what it does.
What you get
A real browser session, driven by an agent
We give an AI shopping agent a realistic task on your store (find a specific product, add it to the cart, reach checkout) and let it work the way agents actually do: reading the page, choosing options, filling forms. It stops before payment.
A narrative friction report
Every step gets a verdict, a timing, and a screenshot. Where the agent hesitated, misread something, or got stuck, the report says exactly what it saw and what it tried.
Prioritized fixes
Blockers first, then friction, then polish, each tied to the step it affected and the change that would have removed it.
Shareable
The report is delivered as a link you can send to your engineering, merchandising, and agency teams.
What a friction log looks like
Example excerpt for a fictional store. Every walk produces a log like this, with screenshots attached to each step.
Step 1 · Search: typed “women’s waterproof trail shoe size 8” into the header search; results loaded in 1.4 s, but the first row was a gift card. Minor.
Step 2 · Product page: price, size chart, and shipping estimate were readable. The “Only 2 left” notice was an image with no alt text, so the agent never saw it. Minor.
Step 3 · Add to cart: the size selector is a custom dropdown with no accessible name; the agent picked the wrong size twice before succeeding. Blocker, 38 s lost.
Step 4 · Cart: a newsletter modal covered the checkout button and needed two dismiss attempts. Friction.
Step 5 · Checkout: guest checkout was available; the address autocomplete silently dropped the apartment number. Friction.
Step 6 · Payment: stopped before payment as agreed. Total 4 min 12 s, 1 blocker, 2 friction points, 2 minor notes.
Pricing
Pricing: coming soon — request a walk and we’ll follow up.
We are running the first walks by hand, one store at a time, and will share pricing with you directly.
Questions
Request an Agent Walk
Tell us the store and what you’d like the agent to try to buy. We’ll confirm by email before anything runs.