Robert Hu · AI Commerce, Strategy & Partnerships
Commerce experience, applied to AI product discovery.
I bring more than 20 years in ecommerce, merchandising, marketplaces, digital transformation, and technology selection. I built RecoScope to investigate a practical question: when a shopper asks AI what to buy, which brands make the answer?
I am exploring AI commerce, strategy, and partnerships roles. This project shows how I frame a commercial problem, build a way to test it, and turn the results into something people can use.
Why I built RecoScope
A shopper's AI answer can name a brand, recommend it first, or mention it only as an alternative. Those are different signals. I wanted a way to examine them across ChatGPT, Claude, Gemini, and Perplexity, using the same shopper situations over time.
RecoScope turns those responses into a working benchmark with dated reports, prompt-level comparisons, and explicit collection and review context. It is a public project that people can inspect, rather than a proposal for a future product.
My contribution
Define the commercial question
I chose product recommendation as the focus, designed a repeatable prompt framework, and separated brand visibility from recommendation position. That makes the benchmark relevant to how commerce teams assess discovery.
Turn answers into usable evidence
I shaped the data model and reporting experience: categories, dated runs, model responses, normalized brand mentions, and findings. Readers can move from a category comparison to the prompts behind it.
Operate and improve the workflow
I run the recurring benchmark, investigate capture failures, check interpretations, and improve the public reports. AI-assisted tools support collection, development, and parsing; I own the scope, measurement rules, and publication decisions.
A decision you can inspect
A mention is not a recommendation.
Counting every brand name as a ranked recommendation would make a clean-looking report with a misleading conclusion. RecoScope keeps mention frequency separate from explicit recommendation rank. Carousel order and unranked alternatives do not become invented ranks.
For a commerce team, that distinction changes the question from “Are we mentioned?” to “Are we selected for this shopper's situation?” It creates a more useful starting point for research and partner discussions, without claiming that visibility caused a sale.
How this work translates to a team
AI commerce
Investigate which brands enter a shopper's consideration set across AI assistants, and how the answer changes with the shopper's needs.
Strategy
Turn an emerging channel into a measurable question, test assumptions, and distinguish observed behavior from claims the evidence cannot support.
Partnerships
Help brands, retailers, and technology teams discuss the same evidence, identify questions worth testing, and define what a useful pilot should measure.
The benchmark measures observed answers, not sales lift or the hidden causes of a recommendation. Published reports carry their own quality and review status; publication does not imply that every run has received human review.
Building an AI commerce team?
I am interested in work connecting AI product discovery, commercial strategy, and partnerships. RecoScope is one example of how I approach that work.