Standing up Chat is mostly Omni-Channel routing plus embedded-widget configuration. The discipline that makes it succeed is bot design, capacity tuning, and reporting cadence.
- Enable Chat (Live Agent or Digital Engagement)
Setup, Chat Settings. Configure deployments, chat buttons, and queue assignments. Service Cloud Voice and Digital Engagement editions ship with Messaging Sessions instead of Live Chat Transcripts.
- Configure Omni-Channel routing
Create Service Channels for Chat, Routing Configurations, and Presence Configurations. Tune capacity (three to five chats per agent is typical) and skill-based routing.
- Build the embedded widget
Configure the Chat embed snippet with branding, pre-chat form fields, and proactive prompt triggers. Drop the snippet into the website pages where chat should appear.
- Front-load with Agentforce or Einstein Bots
For high-volume programs, route incoming chats to a bot first. The bot handles common questions; humans handle the rest. Containment 70 percent or higher is achievable for mature bots.
- Build chat dashboards
Report on chat volume, handle time, CSAT, abandonment, and bot containment. Weekly review against KPIs is what improves chat programs over time.
- Capacity caps too high slow responses and frustrate customers. Three to five concurrent chats per agent is the practical sweet spot.
- Pre-chat forms with too many fields drop conversion. Ask for the minimum needed for routing; collect details after the chat starts.
- Bot fallback paths that miss escalation requests produce dead-end interactions. Test escalation explicitly.
- Live Chat (legacy) and Messaging Sessions (Digital Engagement) use different data models. Reports built on one do not work for the other.