> ## Documentation Index
> Fetch the complete documentation index at: https://docs.wisepilot.app/llms.txt
> Use this file to discover all available pages before exploring further.

# HighLevel Agents

> Configure AI agent prompts, goals, voice, and knowledge bases.

The Agents dashboard lets you manage your HighLevel AI agents — their system prompts, conversation goals, voice settings, and knowledge bases. Better-configured agents book more appointments and qualify more leads.

Navigate to **CRM → Agents** in the sidebar.

## Agent List

The agent list shows all synced HighLevel agents:

| Column                | Description                                    |
| --------------------- | ---------------------------------------------- |
| **Agent name**        | The agent's display name                       |
| **Status**            | Active or paused                               |
| **Performance score** | Composite score based on conversation outcomes |
| **Conversations**     | Total conversation count                       |
| **Booking rate**      | % of conversations that result in appointments |
| **Last active**       | Most recent conversation timestamp             |

Click any agent to open the configuration panel.

## Configuring an Agent

### Prompt Editor

The system prompt is the most important configuration. It tells the agent how to behave, what to say, and what goals to pursue.

**What a good prompt includes:**

```
You are a sales assistant for [Company Name]. Your primary goal is
to qualify leads and book discovery calls.

QUALIFICATION CRITERIA:
- Company size: 50+ employees
- Industry: SaaS, fintech, or healthcare
- Budget: $5K+ monthly marketing spend
- Timeline: Looking to start within 3 months

CONVERSATION FLOW:
1. Greet warmly and ask what brought them to the site
2. Ask about their current challenges with [topic]
3. Qualify using the criteria above
4. If qualified, offer to book a 15-minute discovery call
5. If not qualified, share a relevant resource and offer email follow-up

TONE: Professional but conversational. Use their first name.
Never be pushy. If they say no, respect it gracefully.

ESCALATION: Hand off to a human if:
- They ask about pricing specifics
- They mention a competitor by name
- They express frustration
- The conversation exceeds 10 exchanges without progress
```

**Common prompt mistakes:**

* **Too vague** — "Be helpful and answer questions" gives the agent no direction
* **No qualification criteria** — The agent can't tell good leads from bad ones
* **No escalation rules** — The agent tries to handle everything, including things it shouldn't
* **Too rigid** — Over-scripted prompts make conversations feel robotic

### Goal Settings

Goal fields are part of the agent's entity form, so AI chat can refine them alongside the system prompt using your brand guidelines and ICP context:

* **Agent Personality** — Who the agent is and how it behaves (e.g., "You are a friendly, professional sales assistant")
* **Agent Goal** — Primary objective (e.g., "Book discovery calls with qualified leads")
* **Additional Information** — Business rules, hours, escalation guidelines, examples

<Tip>Use the AI chat to refine goal fields — say "improve the agent personality to be more consultative" and the AI will suggest changes using your brand voice and ICP data.</Tip>

### AI-Powered Prompt Refinement

The agent editor includes an AI chat panel that can analyze and improve your configuration:

1. **Open any agent** from the agent list
2. **Use the AI chat** (right panel) to request improvements
3. **Quick actions** provide one-click prompts for common optimizations:
   * **Objection Handling** — Add responses for common pushback
   * **KB Alignment** — Ensure the prompt references available knowledge base content
   * **Booking Optimization** — Improve appointment conversion language
   * **Tone Calibration** — Match the agent's voice to your brand
4. **Review suggestions** — AI highlights proposed changes in the form fields
5. **Apply** — One click to accept all changes, then **Save** to persist

The AI has full context: your brand guidelines, ICP profiles, linked knowledge bases, and performance metrics. It uses `refine_entity` to suggest targeted improvements rather than rewriting everything.

### Knowledge Base

The knowledge base gives the agent information to reference during conversations. Without it, the agent can only use what's in the system prompt.

**Adding knowledge sources:**

1. Click **Add to Knowledge Base** in the agent config panel
2. Upload documents:
   * **PDFs** — Product brochures, spec sheets, pricing guides
   * **Text files** — FAQ content, competitive positioning, objection-handling scripts
3. Add FAQ entries manually — question/answer pairs for common inquiries
4. Link existing WisePilot content — connect assets from your content library

**What to include:**

* Product features and pricing (if the agent should discuss them)
* FAQ answers for your top 10 customer questions
* Competitive differentiators
* Case study summaries the agent can reference as proof points

<Tip>Start with a focused knowledge base (top 10 FAQs) and expand based on what the [Conversation Intelligence](/reports/conversation-intelligence) report shows your agent struggles with. Adding too much irrelevant knowledge can confuse the agent.</Tip>

## Automated Insights

WisePilot continuously monitors and improves your agents:

* **Nightly sync** — Agent configurations, prompt versions, and conversation outcomes are synced from HighLevel automatically
* **Daily quality scoring** — Every conversation is evaluated against quality rubrics for response relevance, goal progress, and lead handling
* **Daily snapshots** — Per-agent performance metrics (booking rates, conversation volume, eval scores) are aggregated into daily snapshots for trend analysis
* **Weekly prompt analysis** — For agents with 50+ conversations in the past week, WisePilot analyzes conversation patterns and suggests prompt improvements every Monday

See [Automation & Data Freshness](/optimization/automation) for the full schedule.

## Performance Metrics

Each agent's detail view shows:

* **Conversation volume** — Daily/weekly/monthly trends
* **Outcome distribution** — Booked vs. qualified vs. lost (pie chart)
* **Average conversation length** — Number of exchanges per conversation
* **Response quality score** — How relevant and helpful the agent's responses are (scored nightly)
* **Escalation rate** — How often the agent hands off to a human

### Interpreting Performance

| Metric                   | Healthy                    | Needs Attention             |
| ------------------------ | -------------------------- | --------------------------- |
| Booking rate             | > 15%                      | \< 5%                       |
| Qualification rate       | > 40% (booked + qualified) | \< 20%                      |
| Lost rate                | \< 30%                     | > 50%                       |
| Escalation rate          | 5–15%                      | > 25%                       |
| Avg. conversation length | 4–8 exchanges              | > 15 (agent may be looping) |

For deeper analysis, see [Conversation Intelligence](/reports/conversation-intelligence).