Webchat lets your AI agent answer questions from students directly through a chat widget embedded on your website. The agent responds in real time, draws from your knowledge base for accurate answers, and can escalate or follow up by email when needed.
Webchat is configured as a Workflow inside your agent and triggers automatically whenever a chat message is received.
How it works
When a student sends a message through the webchat widget, the agent:
Identifies who it's talking to — using the student's opening message to classify them (prospective student, current student, parent, etc.) and tailor the response accordingly
Queries the knowledge base — searches for relevant information before responding, grounding every factual answer in your knowledge base content
Responds via webchat — delivers answers directly in the chat window
Offers a summary email — at the end of the conversation, the agent can send the student a concise recap of what was discussed
Escalates when needed — if a student's question is outside the agent's scope or triggers an escalation keyword, the agent uses the
escalatetool to route to the appropriate team
Configuration
To set up or review the webchat workflow, go to Workflow in the left sidebar under your agent and open the webchat workflow.
Under the Configure tab, you'll find:
Webchat channel — the identifier for the chat widget connected to this workflow (e.g., student-services-chat). This is set by your CollegeVine team and corresponds to the widget installed on your site.
Actions — the tools the agent can use during a conversation:
knowledge_base— searches your knowledge base for answerswebchat_constituent— sends responses in the chat windowemail_constituent— sends a follow-up or summary emailtext_constituent— sends an SMS follow-upescalate— routes the conversation to a human team
Trigger events — set to Webchat Event Received, which fires on every incoming chat message.
Viewing activity
The Insights tab on the workflow shows a chart of webchat events over time, giving you a quick view of conversation volume and trends.
Testing
Click Test in the top-right corner to run a simulation. Choose a specific constituent to test with, or generate a random one, then start the simulation to see how the agent responds in a live-style interaction.
Going live
The workflow must be set to Live before it will respond to real incoming chats. While in Draft, the agent will not respond to webchat events. Use Pause to temporarily disable responses without archiving the workflow.