How AI Agents and Commerce Intelligence Are Changing Amazon and Walmart

AI is becoming more involved in both customer purchases and the decisions behind marketplace operations. Walmart has announced infrastructure for authorized shopping agents. Amazon is expanding Seller Assistant workflows. Walmart is also developing Scintilla, its supplier insights platform, with a more complete view of first-party retail and Marketplace assortments planned for 2027.

For brands, these developments increase the value of reliable product information and clear operating rules. Gold Compass Commerce recommends strengthening those foundations while testing a limited workflow with an accountable owner. The implications described below are our assessment; platform capabilities and rollout plans remain distinct from their eventual business impact.

Walmart is preparing for agents that can complete purchases

Walmart’s announced Agent Profiles are designed to let customers authorize individual agents and control their purchasing permissions. Its Agent Gateway will provide a dedicated connection for agents, initially through a beta for developers and companies. These capabilities are being introduced in stages. [1]

Our assessment is that brands should prepare for more product evaluation to happen through software. Accurate attributes, compatibility information, availability, and clear descriptions could become increasingly important inputs. That is a readiness priority rather than a published formula for winning agent recommendations.

A pet brand, for example, should make product size, materials, intended use, and relevant limitations consistent across its catalog. An automotive brand should give particular attention to part numbers and supported fitment information. Clear product facts help reduce ambiguity for customers and the systems assisting them.

Amazon Seller Assistant is taking on ongoing work

Amazon has expanded Seller Assistant with continuous workflows and seller-defined guardrails. Sellers can choose recommendation or action workflows, with approval controls and audit trails. Integrations extend these capabilities into Amazon Quick and, in beta, Anthropic’s Claude. [2]

The opportunity is greater monitoring capacity. A small team may be able to investigate more exceptions without manually assembling every report. The business still needs to determine which exceptions matter and what response is appropriate.

For an initial pilot, we recommend monitoring a limited assortment and asking the tool to prepare recommendations for review. Define the objective before setting permissions. Protecting contribution margin, maintaining availability, and maximizing sales can lead to different choices. Record what the system proposed, what the team approved, and the result.

What Walmart Scintilla is and what is changing

Scintilla, previously called Walmart Luminate, is Walmart’s data and shopper insights platform for suppliers. It helps brands analyze their Walmart business and customer behavior. Its roots are in the supplier relationship, where Walmart purchases products for retail sale. [3, 4]

Walmart plans to bring first-party retail and Marketplace assortment insights together more completely in 2027. Its roadmap also includes expanded AI-supported analysis, customizable views, and proactive alerts. The planned combined view should not be presented as universally available today. [3]

For brands selling to Walmart wholesale and through Marketplace, our view is that more connected reporting could improve assortment and availability decisions. However, those models have different economics and operating responsibilities. Keep each model’s contribution and inventory exposure visible beneath the executive summary.

Product information and decision ownership need attention together

These announcements address different users and tasks, but they expose a common management challenge: systems need accurate inputs and a clear definition of the desired outcome.

Assign ownership for product facts, including claims, compatibility, and variant relationships. Establish how changes are approved and distributed. A content team should be able to identify which source governs a disputed specification rather than relying on whichever listing was updated most recently.

Apply the same discipline to internal workflows. Document permitted actions, exceptions that require review, and the person responsible for investigating unexpected results. Sensitive changes involving price, claims, or inventory commitments deserve explicit approval boundaries.

What leaders should do next

Start with an audit of the products that matter most to the business. Resolve incomplete or inconsistent facts before extending automated content or recommendations. Then select one monitoring workflow where faster analysis has measurable value.

Track the accuracy of alerts, time to investigate an exception, approval rates, and the commercial outcome. Review errors and human overrides as evidence about the workflow. Expand only when the team understands both its value and its limits.

Gold Compass Perspective

AI can increase the amount of work a team can handle. The competitive value comes from using that capacity to make stronger commercial decisions. Brands that maintain reliable product information, model-specific economics, and accountable decision owners will be better positioned to evaluate new capabilities as they become available.

How Gold Compass Commerce Can Help

Gold Compass Commerce helps brands connect product content, marketplace operations, reporting, and executive priorities. If you are deciding where AI belongs in your Amazon, Walmart, or DTC business, talk with us about the information and operating controls needed to support a focused pilot.

Discuss Your Commerce Priorities ↗

FAQs

What is Walmart Scintilla?

Scintilla is Walmart’s supplier data and shopper insights platform, formerly Walmart Luminate. It supports analysis of business performance and customer behavior. Available products and access depend on the relevant offering. [3, 4]

Does Scintilla already combine Walmart retail and Marketplace performance?

Walmart announced a more complete combined assortment view for 2027. Brands should distinguish that planned capability from the tools available to them today. [3]

Can AI agents guarantee that a product will be recommended?

No. Complete and accurate product information is a practical readiness measure, not a guarantee of placement or recommendation. Customer intent, available products, and platform behavior can affect outcomes.

Where should a brand begin with AI workflows?

Begin with a specific monitoring task, a limited product group, and human review. Establish the objective, permitted actions, escalation process, and measures of success before expanding permissions.