Agents

Agents are your AI assistants. Think of each agent as a team member with a specific personality, expertise, and way of handling customer conversations.

What is an Agent? #

An agent is a configured AI assistant that:

• Has a specific name and purpose
• Follows instructions you provide
• Speaks in a particular tone or style
• Handles conversations for specific channels or topics
• Uses your documents and playbooks to answer questions

You can have multiple agents, each specialized for different needs.

Creating an Agent #

Step 1: Navigate to Agents #

1. Click Agents in the left sidebar
2. Click Create Agent button

Step 2: Basic Information #

Fill in the essentials:

Agent Name

• Something descriptive: "Customer Support Agent", "Sales Assistant", "Technical Support Bot"
• Visible to your team (not to customers)

Description (Optional but recommended)

• What does this agent handle?
• Example: "Handles order questions, shipping inquiries, and basic product information"

Step 3: Configure Behavior #

Instructions

Tell your agent how to behave. Write in plain English:

You are a helpful customer support agent for an online store.

Your job is to:
- Answer questions about orders, shipping, and products
- Be friendly and patient
- Always check our documentation before answering
- Escalate to a human if you're not confident

Important guidelines:
- Never make promises about shipping dates
- Always verify order numbers before looking them up
- If someone is frustrated, offer to connect them with a manager

Tone

Choose how your agent should communicate:

Things to Avoid (Optional)

List topics or phrases your agent should never use:

- Don't promise specific dates without checking systems
- Avoid technical jargon with non-technical customers
- Never say "that's not possible" without offering alternatives
- Don't make pricing decisions without approval

Step 4: Initial Greeting #

Set a welcoming first message customers see:

Examples:

E-commerce:

Hi! Welcome to [Store Name]! I'm here to help with orders,
products, and shipping. What can I help you with today?

SaaS Support:

Hello! I'm your [Product Name] assistant. I can help with
account questions, troubleshooting, or feature guidance.
How can I assist you?

General Support:

Hi there! Thanks for reaching out. I'm here to help!
What brings you in today?

Step 5: Enable and Save #

• Toggle Enabled to activate your agent
• Click Save Agent

Your agent is now ready to start handling conversations!

Editing an Existing Agent #

Need to make changes? Easy!

1. Go to Agents in the sidebar
2. Click on the agent you want to edit
3. Make your changes
4. Click Update Agent

Changes take effect immediately for new conversations.

Agent Settings #

Test Mode #

Enable test mode when you want to review every response before it's sent.

When test mode is ON:

• Agent generates responses as normal
• You see responses before they're sent to customers
• You can approve, edit, or reject each message
• Perfect for training and quality assurance

How to enable:

1. Open your agent settings
2. Toggle Test Mode ON
3. All conversations with this agent now require approval

Human Handoff Instructions #

Tell your agent exactly how to handle escalations:

When Human is Available:

Let the customer know that you'll connect them with a specialist
who can help. Transfer the conversation and provide a summary of
what was discussed.

When Human is NOT Available:

Apologize and let the customer know you'll have someone reach out
within 2 business hours. Collect their email if you don't have it.
Ask if there's anything else you can help with while they wait.

Trigger Conditions #

Set when this agent should be assigned to conversations:

By Channel:

• Assign this agent to WhatsApp conversations
• Different agent for website chat
• Another for email

By Topic:

• Keywords: "order", "shipping", "tracking" → Order Support Agent
• Keywords: "bug", "error", "not working" → Technical Support Agent
• Keywords: "price", "plan", "upgrade" → Sales Agent

By Time:

• Business hours → Human + AI assist
• After hours → AI only

Multiple Agents Strategy #

When to Use Multiple Agents #

Good reasons:

• Different expertise areas (sales vs support)
• Different channels (WhatsApp vs email)
• Different languages or regions
• Different customer tiers (free vs premium)

Not necessary for:

• Slight tone differences
• Similar topics
• Same channel

Agent Specialization Examples #

Example 1: E-commerce Store

Support Agent

• Handles: Order questions, shipping, returns
• Tone: Friendly and helpful
• Channels: All

Sales Agent

• Handles: Product questions, recommendations
• Tone: Enthusiastic and knowledgeable
• Channels: Website chat only

Example 2: SaaS Company

First-line Support

• Handles: Account questions, basic troubleshooting
• Escalates: Technical bugs, billing issues
• Channels: In-app chat, email

Technical Support

• Handles: Complex technical issues, API questions
• Tone: Professional and detailed
• Channels: Email, priority queue

Agent Performance #

Monitoring Your Agents #

Go to Analytics → Agents to see:

• Conversations handled - Total volume per agent
• Resolution rate - How many issues they solved
• Average response time - Speed of responses
• Escalation rate - How often they need human help
• Customer satisfaction - Ratings from customers

Improving Agent Performance #

If resolution rate is low:

1. Review conversations the agent struggled with
2. Add more training documents on those topics
3. Update agent instructions with examples
4. Create specific playbooks for common issues

If escalation rate is high:

1. Check why conversations are escalating
2. Give agent more confidence in instructions
3. Add "decision trees" in playbooks
4. Expand knowledge base coverage

If responses are too slow:

1. Simplify agent instructions
2. Organize documents better
3. Reduce unnecessary playbook steps

If customer ratings are low:

1. Review exact conversations with poor ratings
2. Adjust tone settings
3. Add more empathy to instructions
4. Train on better response examples

Best Practices #

Writing Great Instructions #

Do:

• ✅ Be specific and clear
• ✅ Use examples
• ✅ Include what TO do, not just what NOT to do
• ✅ Think like you're training a new employee

Don't:

• ❌ Be vague ("be helpful")
• ❌ Use technical jargon
• ❌ Write a novel (keep it under 500 words)
• ❌ Contradict yourself

Starting Simple #

For your first agent:

1. Start with basic, clear instructions
2. Test in playground mode
3. Handle one main purpose well
4. Expand gradually based on real conversations

Regular Maintenance #

Set a reminder to review your agents:

Weekly:

• Check performance metrics
• Review a few conversations
• Update instructions if needed

Monthly:

• Deep dive into problem areas
• Add new capabilities
• Retire unused agents

Common Questions #

Can I have multiple agents responding to the same customer? #

No, Hay assigns one agent per conversation to maintain consistency. You can manually reassign if needed.

What happens if I disable an agent? #

New conversations won't be assigned to this agent. Existing conversations continue with the same agent.

Can agents learn on their own? #

Agents improve based on the documents you add and playbooks you create. They don't automatically "learn" without your guidance.

How many agents should I have? #

Most businesses start with 1-2 agents. You only need more if you have truly different use cases or channels.

Can I clone an agent? #

Yes! When creating a new agent, you can start with settings from an existing agent, then customize.

Next Steps #

• Set up Playbooks to give your agent step-by-step workflows
• Upload Documents to train your agent
• Review Conversations to see your agent in action