Assemble coordinated merchandise recommendations

What the source reports

OpenAI reports that Fanatics uses enriched product data, team information and inventory to assemble coordinated, in-stock looks. Merchandising rules constrain the selections, which refresh daily. Engineers built recommendation tools with Codex; Fanatics reviewed results, supplied feedback and approved the initial product before production rollout. The account reports 10% higher engagement after testing about 30,000 MLB products.

Reported technology: Fanatics with OpenAI; OpenAI models and Codex. Scope and outcomes are source-reported.

What the enterprise environment would need

The following is a proposed project design for an enterprise doing similar work. Published details are marked; the remaining systems, data and scenarios would be agreed for your project.

Systems and evidence

Proposed: use a fictional merchandise catalogue with versioned inventory, team identifiers and explicit compatibility rules.

People and agents

Proposed: an agent drafts coordinated looks; a merchandiser checks compatibility and exceptions, while an engineer owns the publishing integration.

State and permissions

Proposed: retain recommendation inputs, rule versions, stock timestamps and approvals. Keep the practice system separate from live merchandising.

A proposed simulation

An illustrative practice design derived from this work, not an executed assessment.

Proposed: assemble an approved fictional game-day look while respecting stock, team and merchandising constraints.

Review gate: An independent reviewer can reproduce each recommendation from its recorded inputs, verify every constraint and identify items requiring human judgment.

Challenge the result: Inventory changes after a look is drafted. The agent must detect the stale stock record and withhold the unavailable item before publication.

What needs checking

• Proposed: independently verify team compatibility and stock availability for every recommended item.
• Proposed: reject recommendations that violate a merchandising rule, even if visually plausible.
• Proposed: compare engagement and availability failures against a predefined baseline without treating clicks as purchases.

Review environment fidelity, human capability and agent performance separately. The enterprise’s accountable owners decide acceptance of the shared workflow and any operating use.

Shape this around your enterprise

Bring a similar workflow and the people responsible for it. Together, define the records and systems to use, the result your team will build or configure, and how its owners will check it. Nuvepro scopes the simulation, environment and preparation. Agree any existing-agent evaluation, implementation assistance and production responsibilities separately.

Sources and limitations

Curation review: 27 October 2026. This is the date this source example was reviewed, separate from source publication dates.

• This is a provider-published customer account, not an independent study.
• The engagement claim does not establish revenue uplift; the article does not provide an evaluation duration, error rate or detailed experimental design.
• The proposed environment and checks are illustrative practice designs, not a delivered Nuvepro product or a replica of Fanatics systems.

1. How Fanatics and OpenAI streamline the fan experienceSource publication: 2026-10-26