Crocs replaced its underperforming legacy chatbot with Rivet, a single AI agent built on Salesforce Agentforce that handles product discovery, sizing, order tracking, returns and checkout within one continuous conversation. The shift was driven by operational necessity: the company's previous chatbot routed 80% of interactions to human agents, whilst its service team fielded 178,000 inquiries monthly—over half of which were routine order status requests. Rather than deploying disconnected agents for separate tasks, Crocs opted for a unified interface backed by a coordinator that manages multiple specialized subagents, a design choice that reduced average handle time by 50% and halved redundant customer profiles. The architecture proved flexible enough that Crocs extended Rivet to phone support in seven weeks by reusing the same underlying actions and instructions, demonstrating how orchestrated agent systems can scale across channels faster than traditional multi-tool implementations.
The deployment reveals a critical tension in agentic CX: retailers are racing to invest in AI shopping assistants—37% plan significant investment over three years—yet consumer willingness to delegate remains conditional and task-dependent. Whilst 64% of consumers expect to use AI shopping agents within two years and 50% have already made purchases with AI assistance, only 24% would permit an agent to complete payment autonomously. This gap between research delegation and transaction authority means checkout represents Rivet's highest-friction use case, where trust must be earned through demonstrated reliability rather than assumed. For CX teams already running Agentforce or evaluating similar platforms, the Crocs case illustrates that orchestration architecture matters as much as individual agent capability—the ability to add specialized subagents without rebuilding the entire system directly impacts deployment velocity and operational resilience.
The broader implication cuts across vendor strategy and team readiness. Crocs's five-month implementation timeline and measurable efficiency gains (50% reduction in handle time) will likely accelerate enterprise adoption of orchestrated agent systems, potentially shifting budget allocation away from single-purpose chatbots and towards platforms that support coordinator-based architectures. However, the persistent consumer hesitation around payment delegation suggests that teams cannot simply automate their way to resolution—Rivet's success will ultimately depend on whether it can convert the 76% of Americans currently unwilling to delegate checkout, a challenge that sits at the intersection of AI capability and human psychology rather than pure technical implementation.
Crocs Puts One AI Agent in Charge of Sales and Service PYMNTS.com