Crocs Puts One AI Agent in Charge of Sales and Service

A shopper looking for a birthday gift for a nephew opens a chat on Crocs.com, asks for ideas, checks whether a size will fit and wants to know how to return it if it doesn’t. Until this year, that meant bouncing between a product page, an FAQ and a support queue. Crocs now wants the whole exchange to happen in one conversation with Rivet, its new artificial intelligence (AI) agent.

Before Rivet, the footwear company’s service team was handling 178,000 inquiries a month across voice, chat and email, Salesforce said in a customer case study. More than half were “Where is my order?” requests. The rest covered returns, refunds, sizing and policy questions. Crocs had a self-service chatbot, but its limits meant 80% of those conversations were routed to a service rep.

Why Crocs Chose One Agent Over Multiple

Crocs sold 129 million pairs of shoes in 2025 and wanted its digital experience to grow with that volume, Salesforce wrote. The company rejected both a basic FAQ bot and a set of disconnected agents for different tasks. It wanted a single assistant that could guide product discovery, help with fit, resolve service questions and know when to bring in a person, all in one continuous chat.

“Having multiple agents could add friction for the consumer,” said Feliz Papich, senior vice president of digital technology, experience and insights at Crocs.

Crocs built Rivet on Salesforce’s Agentforce platform with consulting partner Rosetree Solutions. The three companies took it from concept to customer-ready in five months. The first release covers five jobs: product discovery, FAQs, order tracking, returns and conversational checkout.

Behind Rivet, a Team of Specialized Agents

Under the hood, Rivet is several agents. “We designed the first release of Rivet to solve focused use cases so we could learn from real interactions before expanding,” Rosetree managing partner Craig Rosenbaum said in the Salesforce study. A single orchestrator coordinates specialized shopper and service agents, so Crocs can add a new use case by adding a new subagent rather than rebuilding. The service agent pulls real-time order status, tracking and return eligibility from Crocs’s IBM Sterling order management system.

That structure let Crocs extend Rivet to phone calls in seven weeks by reusing the same actions and instructions built for chat. Before each release, the company ran hundreds of conversation scenarios through an automated testing tool that scores responses against brand guidelines.

Crocs reported 50% fewer redundant customer profiles and 50% lower average handle time from unifying service data, figures that came from the company and its vendor rather than an external audit. Rivet’s own results are still to come. Crocs expects Rivet to lift the share of conversations resolved without a human, along with search success, click-through rate, average order value and items per order.

Shoppers Still Hold the Buy Button

Crocs is building what retailers say they want next. Thirty-seven percent of retailers plan to invest in AI shopping assistants over the next three years, more than any other digital capability, PYMNTS Intelligence found in “Global Digital Shopping Index: The AI-Powered Shopper Has Arrived,” a report commissioned by Visa Acceptance Solutions, Shoppers appear ready to follow. The same report found 64% of consumers expect to use AI shopping agents within two years.

Checkout is the one job on Rivet’s list that shoppers are least ready to give up. Half of Americans say they’ve made a retail purchase with AI’s help, but only 24% would let an agent shop and pay for them, PYMNTS Intelligence reported in “Will the 2026 Shopping Season Go Agentic?” Consumers are comfortable delegating research, the report said, and far less comfortable delegating payment choice or an irreversible decision. That makes checkout the job where Rivet will have to earn the most trust.