Wayfair is using OpenAI models to improve product catalogue accuracy and speed up support workflows at scale. By embedding models into supplier support routing and catalogue attribute systems, Wayfair can automate triage, reduce manual effort, and correct product attributes across millions of items—while keeping high-impact actions governed through approvals and auditing.

Retail operations are full of work that’s both repetitive and high-stakes: product attributes must be correct across millions of listings, and support teams need to resolve issues quickly without losing context.

Wayfair has taken a “build it into the workflow” approach to generative AI. Instead of treating models as a separate experiment, Wayfair embedded OpenAI models into core operational systems—focusing first on two areas where complexity and scale are unavoidable: supplier support and catalogue quality.

What Wayfair is doing with OpenAI

Wayfair’s public case study describes OpenAI models being integrated into critical internal workflows to:

• Improve catalogue quality by classifying and correcting product attributes consistently

• Automate support ticket triage so requests route faster and resolutions accelerate

The logic is simple: when your catalogue has tens of millions of products, even “small” inconsistencies become a major operational cost.

Why catalogue attribute accuracy matters

Product attributes sit underneath almost everything:

• search and filtering

• recommendations

• merchandising

• returns and customer satisfaction

If attributes are wrong (dimensions, materials, colour families, compatibility), customers lose trust and operations absorb the fallout.

LLMs can help here by standardising how attributes are interpreted and mapped—especially when supplier data arrives in inconsistent formats.

Faster ticket handling through triage automation

Support workflows often slow down at the start: categorising the ticket, routing to the right queue, identifying the required data, and drafting a first response.

A practical agentic pattern looks like this:

1. Classify the request (issue type, urgency, required policy)

2. Route to the correct queue or owner

3. Enrich with missing fields (what’s needed to resolve)

4. Draft a proposed response or next-step plan

5. Approve and send (human-in-the-loop for anything risky)

Wayfair’s engineering learnings highlight that even “triage only” automations can create meaningful speed improvements because they remove the constant switching between systems and categories.

How to replicate this pattern in your organisation

If you’re looking at similar use cases—retail, marketplaces, B2B catalogues, or any environment with large-scale product data—the safest path is structured.

Step 1: Start with a bounded pilot

Pick one catalogue domain (e.g., furniture materials) or one support category (e.g., delivery exceptions).

Step 2: Define the source of truth

LLMs can normalise, but they shouldn’t invent. Establish which systems and fields are authoritative.

Step 3: Design for governance

• keep write actions behind explicit tools

• add approval steps for changes to customer-facing records

• log outputs and decisions for audit and debugging

Step 4: Measure impact in business terms

For catalogue:

• attribute accuracy against a labelled set

• reduction in manual QA effort

• improvements to search/filter performance

For support:

• time to first action

• routing accuracy

• resolution time and re-open rates

Where Generation Digital can help

These are the kinds of workflows where “AI” can either create chaos or become a dependable operational layer.

Generation Digital can help you:

• identify high-impact catalogue and support automations

• design safe agent workflows (tools, approvals, auditing)

• implement evaluation so quality improves over time

• align teams on governance and adoption

Summary

Wayfair’s integration of OpenAI models shows what “AI in production” looks like in retail: embed models into the workflows that create operational drag—catalogue accuracy and ticket triage—and measure outcomes in speed, quality, and reduced manual effort.

Next steps: If you want to apply the same approach to your catalogue, support workflows, or internal operations, speak with Generation Digital: https://www.gend.co/contact

FAQs

Q1: How does OpenAI improve Wayfair’s product catalogue?
By helping classify and correct product attributes consistently at scale, reducing inconsistency and manual remediation.

Q2: What impact does this have on support?
Automation can categorise and route tickets faster, enrich requests with the right context, and draft responses—so teams spend less time on sorting and more time resolving.

Q3: Is the automation scalable?
Yes. These workflows are designed for high volume: model-driven classification and enrichment can run across large catalogues and constant ticket inflow.

Q4: What’s the biggest risk in catalogue automation?
Uncontrolled write access. The safe approach is “compute then propose”, with approvals and auditing for any customer-facing change.

Q5: What should we pilot first?
Start with one category or one ticket type, build an evaluation set, and prove accuracy before expanding to broader automation.