Case Triage Service Agent
Automotive
250+ employees
BigSolve built an Agentforce service agent that has classified 50,000+ inbound cases at 90%+ accuracy — replacing manual triage with an LLM agent that acts autonomously.
The Challenge
An online used-car buyer's customer support team handles a steady volume of inbound Email and SMS cases — most of them offer cancellations, the rest a mix of status inquiries and auto-reply noise. Routing each case correctly required human triage on every message.
Before partnering with BigSolve, classification ran on a rigid Flow with hard-coded keyword matching, while resolution depended on agents reading every case to decide whether to cancel an offer, escalate for review, or close it out.
They needed an AI-native triage layer inside Salesforce — one that could read free-text customer messages, classify intent reliably, and act on the routine paths without pulling an agent in.
.avif)
Summary
An Agentforce service agent that classifies inbound cases and executes the right action — cancel, review, or close — autonomously, with its reasoning posted to each Case's Chatter feed.
Differentiator
Measured rollout — ran the classifier headlessly across thousands of real cases until accuracy crossed 90%, then promoted to a full service agent only when the data justified it.
Process Overview
Mapped the support team's inbound case patterns and the legacy triage flow's failure modes.
Built a Flow + Prompt Builder classifier and ran it headlessly to measure real accuracy.
Iterated the prompt template until classification crossed the 90%+ benchmark.
Promoted the proven classifier into an Agentforce service agent with autonomous actions.
Get Started with an Expert-Led Discovery
