Agent Groups
Agent Groups
An agent group is a team of specialist agents that collaborate on conversations. When a message comes in, every agent in the group contributes their perspective, but only the lead agent's response is sent to the visitor. The other agents act as consultants, informing the lead's response with their expertise.
When to Use a Group vs a Single Agent
Use a single agent when:
- The conversation covers one topic (just support, just qualification, just onboarding)
- One skill set is enough to handle the request
- You want to keep things simple while getting started
Use a group when:
- Conversations span multiple areas of expertise
- You want richer responses that draw on different perspectives
- A single agent keeps getting things wrong because it is trying to do too much
For example, a pre-sales conversation might need someone who can qualify the lead, someone who can answer technical questions, and someone who can talk about pricing and next steps. One agent trying to do all three often gives shallow answers. A group of three specialists, each contributing what they know best, gives a much stronger response.
Creating a Group
- Go to AI > Agents
- Click Create Group
- Give the group a name (e.g., "Pre-Sales Team")
- Add agents as members
- Set one agent as the lead
- Save and activate
The Lead Agent
The lead agent is the one whose response gets sent to the visitor. Choose the agent whose role is closest to the primary purpose of the group.
For a pre-sales group, the lead might be your qualification agent. For a success group, the lead might be your account manager agent.
Consulting Agents
Every other agent in the group acts as a consultant. They analyze the conversation and share their insights with the lead agent before it responds. The visitor never sees the consulting agents directly - their expertise is woven into the lead's response.
This means you can add a technical expert to your sales group without worrying about the response sounding too technical. The lead agent takes the technical input and translates it into language that fits the conversation.
Example Group Configurations
Pre-Sales Team
A group for handling inbound prospects who are evaluating your product.
| Agent | Role | What it contributes |
|---|---|---|
| Lead Qualifier | Lead | Asks qualifying questions, captures details, steers the conversation |
| Technical Expert | Consultant | Answers technical questions, explains integrations, clarifies capabilities |
| Solutions Consultant | Consultant | Recommends the right plan, handles pricing objections, suggests next steps |
When to use this: Website visitors asking "does your product do X?" or "how much does it cost?" or "can it integrate with Y?"
Success Team
A group for handling post-sale customer conversations.
| Agent | Role | What it contributes |
|---|---|---|
| Onboarding Specialist | Lead | Guides setup, explains features, answers getting-started questions |
| Account Manager | Consultant | Surfaces usage data, identifies expansion opportunities, flags risk |
| Upsell Closer | Consultant | Recommends upgrades when usage suggests a better-fit plan |
When to use this: New customers going through setup, or existing customers with account questions.
Support and Sales Hybrid
A group for handling mixed inbound where some visitors need help and others want to buy.
| Agent | Role | What it contributes |
|---|---|---|
| Customer Service | Lead | Triages the request, answers support questions, resolves issues |
| Lead Qualifier | Consultant | Identifies buying signals, captures details if the visitor is a prospect |
| Data Entry | Consultant | Extracts contact and company information for CRM updates |
When to use this: When your chat widget handles both support and sales inquiries on the same page.
How Routing Works
When you have multiple agents or groups connected to the same source (e.g., your website chat widget), Outrun automatically creates a router.
What the Router Does
The router reads each incoming message and decides which agent or group should handle it. It classifies the intent of the message - is this a support question, a sales inquiry, a technical question? - and sends it to the best match.
You do not need to configure the router manually. It is created automatically when two or more agents or groups share a source.
Routing Criteria
You can add routing criteria to fine-tune which conversations go where. Criteria can filter based on:
- Visitor attributes - identified vs anonymous, company size, industry
- Segment - which visitor segment they belong to
- Page context - which page the conversation started on
- Message content - keywords or intent signals in the message
For example, you might route identified enterprise visitors to your pre-sales group and anonymous visitors to your lead qualification agent.
How Routing Feels to the Visitor
Routing is invisible to the visitor. They send a message and get a response. They do not see agent names, group names, or routing decisions. The experience is a single, coherent conversation regardless of how many agents are involved behind the scenes.
Tips
- Start with a single group. Get one group working well before creating more. Test it with real conversations and adjust the membership based on what you see.
- Choose your lead carefully. The lead agent sets the tone. If your group is for sales, make the sales agent the lead. If it is for support, make the support agent the lead.
- Review group responses with HITL. Enable human-in-the-loop review on the group when you first set it up. This lets you see how the consulting agents influence the lead's response, and whether the result is better than a single agent alone.
- Keep groups small. Two to three agents per group is the sweet spot. More than that and the consulting inputs can conflict, making the lead's response less focused.
- Use routing criteria sparingly at first. The auto-router works well out of the box. Add criteria only when you see conversations being routed to the wrong place.
