Testing AI Agents in BoldDesk is essential to ensure they respond accurately and consistently before you go live. Use the built‑in Preview panel to simulate customer questions, validate intent understanding, and verify tool/API behaviours—without leaving the configuration screen. This is also the best place to run multi‑agent checks across roles (e.g., billing, onboarding, tech support) and brands.

Prerequisites

For an Agent to customize and test an AI agent in BoldDesk, their role must have the permission to Manage AI Agents.

Steps to Test an AI Agent in BoldDesk

1) Open the AI Agents list

1. In the left sidebar, select AI.

2. Choose AI Agents to open the list of agents.

   

2) Pick the agent you want to test

• On the AI Agents page, click the agent card.
  You can use the tabs (All | Published | Draft) and the brand filter (top‑right) to find the correct agent.

  

3) Test from the Configure tab’s Preview panel

1. On the agent page, ensure you’re on Configure.

2. In the left rail, switch between configuration categories as needed:

   • Persona
   • Guidance
   • Knowledge Sources
   • AI Tools
     • API
     • MCP Server

3. Use the Preview panel on the far right to test at any time (it is visible in every Configure category).

4. Type a query in the “Type your message here…” box and review the response.

5. If you edited settings on the page, click Update first to apply changes, then re‑test.

   

What to Validate While Testing

A) Understanding & intent recognition

• Ask varied, realistic customer questions across topics.
• Check if the agent’s response is on-topic and grounded in the right Knowledge Sources.
• Verify routing/role alignment (e.g., a billing query doesn’t answer with onboarding content).

B) Tools & actions (updated names)

• AI Tools: Trigger tool behaviours with the phrases you expect customers to use (e.g., “Update my billing address”). Confirm the agent chooses the correct tool and returns the right outcome.
• API / MCP Server: For API-backed tasks, test prompts that should call your API or MCP tool endpoints, and verify data is returned/acted on correctly.

C) Tone & policy adherence

• Review clarity, length, and brand voice set in Persona and Guidance.
• Confirm the agent avoids off-limits content and provides safe fallbacks.

D) Iterate, then publish

• Tweak Guidance, Knowledge Sources, or Tools/API based on results.
• Click Update to save, re‑test, and when satisfied, select Publish Agent (top‑right).

Multi‑Agent & Multi‑Brand Testing

• From the AI Agents list, use the brand filter to switch between brands and test each agent in its designated scope.
• Test each agent against the topics it owns (e.g., Billing Agent vs. Tech Support Agent) to ensure role isolation and correct responses per domain.
• If agents are mapped to different channels, mirror likely channel queries during Preview testing to anticipate live behaviour.

Frequently Asked Questions (FAQs)

1. Can I test AI Agents in BoldDesk before deploying them?
Yes. Use the Preview panel on the right side of the Configure tab to simulate conversations.

2. Do I need special permissions to test AI Agents in BoldDesk?
Yes. The Manage AI Agents permission is required.

3. Can I test AI Agents tools and integrations in BoldDesk?
Yes. Use the Preview panel to trigger AI Tools and verify API/MCP behaviours with realistic customer phrases.

Related Article

Unveiling the BoldDesk AI Agent: Your 24/7 Virtual Support Assistant