Simply AskingDocs

Building Your First AI Agent

Create an AI agent powered by your knowledge base, configure its persona and grounding, and test it with real questions before publishing.

This guide is for admins and owners who want to create AI agents that answer questions using your organization's knowledge base. No coding is required. You will configure your agent through a step-by-step wizard.

What you will accomplish: Create an AI agent powered by your knowledge base, configure its persona and grounding, and test it with real questions before publishing.

1

Navigate to Agents

Click AI in the header, then select Agents. Click Create Agent to start the wizard.

2

Name and Describe Your Agent

Give your agent a clear, descriptive name (e.g., "Sales Assistant", "IT Help Desk", "Onboarding Guide"). Write a short description of what it does so your team knows when to use it.

3

Set the Persona

Define how the agent should behave: its tone, area of expertise, and response style. For example: "You are a friendly IT support specialist who helps employees troubleshoot common technical issues. Keep answers concise and include relevant links."

4

Choose a Grounding Mode

Select the grounding mode that controls how strictly the agent sticks to your knowledge base.

GroundedAgent only uses information from your knowledge base. Most restrictive, best for compliance and regulated environments.
BalancedAgent primarily uses your KB but can provide general context when needed. Good for most use cases.
CreativeAgent uses your KB as a starting point but can expand on topics. Best for brainstorming and exploratory conversations.
5

Select Knowledge Scope

Choose which documents and departments the agent can access. A narrower scope produces more focused, accurate answers. You can always expand the scope later as your needs grow.

6

Test Your Agent

Ask the agent questions you expect real users to ask. Test a range of scenarios: common questions, edge cases, and topics outside its scope. Check that answers are grounded in your knowledge base and that the agent handles unknown topics gracefully.

7

Review Response Quality

Evaluate the agent's responses. Are answers accurate? Are sources cited correctly? Does the tone match your persona? If anything needs adjustment, go back and refine the persona or knowledge scope before publishing.

8

Publish the Agent

When you are satisfied with the agent's responses, toggle its status from Draft to Active. The agent is now available to your team and ready to answer questions.

Pro Tip: Start with a narrow knowledge scope and expand over time. An agent that answers 10 questions perfectly is more valuable than one that answers 100 questions poorly. Test with real questions your users actually ask, not hypothetical ones.

Your agent is live and ready to answer questions. Next, learn how to deploy it to external channels, monitor its performance, and fine-tune its behavior based on real conversations.

Was this helpful?