Agent Assist AI Employee
The Agent Assist AI Employee is built from the Agent Assist curated template (category: Support). It is a chat AI Employee (AIE) that works behind your human support agents: it classifies an incoming query by category and urgency, then searches your internal knowledge for the most relevant resolution steps. Rather than replying to the customer directly, it surfaces suggested resolutions inline so the agent can resolve faster. It cuts average handle time and helps new agents ramp up with contextual guidance.
Workflow
The template ships a two-agent workflow that runs left to right:
| Node | Agent type | What it does |
|---|---|---|
| Classify Query | intent_classification | Sorts the support query into a category and assigns an urgency level. Ships with four categories: billing, technical, account, general. |
| Find Resolution | search_respond | Searches your internal documentation for resolution steps matching the query and classification, and returns the most relevant resolution paths with citations. |
A start node feeds the support query into the graph, and a publish node exposes the suggested resolution. The edges run start → classify → search → publish.
The Classify Query node ships with its intent categories pre-filled in the agent's type_config. For example:
{
"intents": [
{ "label": "billing", "description": "Billing, payment, or subscription queries",
"examples": ["Refund request", "Charge dispute"] },
{ "label": "technical", "description": "Product bugs, errors, or feature questions",
"examples": ["App not loading", "Integration not working"] },
{ "label": "account", "description": "Account access, profile, or security",
"examples": ["Password reset", "Locked out of account"] },
{ "label": "general", "description": "General inquiries",
"examples": ["Hours of operation", "Return policy"] }
]
}Starting configuration
The template clones in with this configuration, which you can change in the AI Employee builder:
- Model selection — Let Fusion Pick (EmaFusion™ chooses the model per request).
- Optimization priority — Balanced.
- Conversation context — enabled, so the agent retains context across turns.
- Feedback — enabled, with thumbs-up / thumbs-down reasons and free-text comments.
- Welcome message — "Hello! I'm here to help you find the best resolution for your support query."
- Conversation starters — two buttons: "Troubleshoot an issue", "Find a solution".
Recommended data sources: connect your internal knowledge base, troubleshooting guides, and product documentation so the agent can suggest accurate resolutions.
How to use
After you select Use template and the AI Employee opens in the builder:
- Connect your internal knowledge base. Add your runbooks, resolution guides, and past ticket summaries to the Find Resolution node. See Knowledge bases.
- Customize the classification categories. Edit the categories and urgency levels in the Classify Query node to match your team's ticket taxonomy.
- Publish and integrate. Publish the workflow, then connect it to your ticketing platform so agents see suggestions inline. See Channels and deployment.
Agent-facing, not customer-facing. Agent Assist is designed to suggest resolutions to your human agents, not to reply to customers directly. For an AI Employee that responds to customers on its own, start from the Customer Support Chat template instead.
Example queries
- "Customer says their invoice was charged twice this month."
- "User can't connect the Slack integration after re-authenticating."
- "Account is locked after too many failed login attempts."
Related
- Agent reference — the Intent Classifier and Search & Respond agent types.
- Knowledge bases — connecting your internal resolution content.
- Templates overview — the full curated template list and the clone flow.