Preference Collection Limits Agentic Shopping

The key constraint in agentic shopping is not payment, it is preference collection. Most purchases are a fast sequence of tiny choices about quantity, material, shipping speed, merchant trust, returns, and budget. Vetted is built around helping a shopper make those choices step by step, while generic assistants still struggle to infer them reliably from sparse context and inconsistent merchant data.

• Vetted already sees shopping as a multi turn workflow, not a single prompt. Users start with broad searches like electric scooter, then refine through back and forth on use case, budget, comparisons, and price history. That is exactly the kind of decision path a one shot agent skips.
  1 sacra
• This is why research and checkout are separating into different hard problems. Upstream, the system has to narrow choices and build trust. Downstream, it has to sync inventory, pricing, shipping, fraud checks, and order approval. Stripe and OpenAI are solving the checkout rail, but that still does not solve product fit.
  1 sacra 3 stripe 4 stripe
• The practical winner is likely a hybrid interface. OpenAI added shopping results in ChatGPT search in April 2025, and Perplexity offers Instant Buy, but both still depend on structured merchant data and limited product context. In many categories, a visual choice set still works better than asking an agent to guess the right variant.
  1 sacra 2 openai 5 perplexity

The next phase of commerce will combine conversational guidance with structured product selection. The companies that win will be the ones that can ask the missing questions, render the right options, and connect that flow to reliable checkout infrastructure. That favors specialized shopping layers like Vetted, even as generic assistants get better at completing transactions.