How OpenTable Uses Agentforce to Resolve 70% of Customer Inquiries
OpenTable deployed Salesforce Agentforce across restaurant and diner support channels, autonomously resolving 70% of inquiries and handling 73% of restaurant web queries within 3 weeks — projected to resolve 180,000 cases annually.
Impact
70%
Diner and restaurant inquiries resolved autonomously
73%
Restaurant web queries handled by agent
40%
Improvement over prior chatbot
180,000
Projected annual cases resolved
3 weeks
Time to deploy restaurant agent
Challenge
OpenTable's customer service teams were consumed by high volumes of repetitive low-complexity inquiries across both restaurant and diner support channels, limiting capacity for complex issues.
Solution
Salesforce Agentforce deployed as autonomous AI agents for restaurant and diner support, grounded on 1,500 Service Cloud knowledge articles and governed by a real-time deflection scoring system to determine when to resolve, escalate, or hand off to humans.
Tools & Technologies
What Leaders Say
“Agentforce handles simple inquiries automatically, so our people can focus on delivering superior service.”
Sign up to read complete case studies, access detailed metrics, and unlock all use cases.
Full Story
OpenTable connects 60,000 restaurants with 1.9 billion diner reservations annually, operating a dual-sided marketplace that requires customer support for both restaurant owners managing their operations and diners searching for and booking tables. As scale grew, the burden on customer service teams from repetitive, low-complexity inquiries — opening hours, menus, addresses — consumed capacity that could be better directed at complex issues.
OpenTable worked with Salesforce to deploy Agentforce as an autonomous AI agent layer across both sides of its marketplace. The implementation grounded agents on a knowledge base of 1,500 articles stored in Salesforce Service Cloud, with Salesforce Data Cloud providing retrieval-augmented generation to surface precise, contextual answers. A key innovation was a real-time deflection scoring system: the Atlas Reasoning Engine dynamically scores each conversation based on sentiment and intent — a request for help scores 5, a request for a human scores 10, extreme frustration scores up to 20 — and Agentforce uses that score to decide whether to continue handling, create a case, or escalate to a human agent with a full transcript.
The restaurant agent went live in just 3 weeks on the Contact Us page. The diner agent followed shortly after. Both were designed to handle common high-volume queries automatically while ensuring complex or emotionally charged interactions reached human agents with full context already in hand.
Results were rapid: within weeks of launch, 73% of restaurant web queries were being handled by the agent, and 70% of diner and restaurant inquiries were resolved autonomously — a 40% improvement over OpenTable's prior chatbot. The system is projected to resolve 180,000 cases per year, freeing service teams to focus on the 30% of interactions that genuinely require human judgement.