Perform Research with Knowledge Agents

Guru delivers AI-powered research anchored in your company’s institutional knowledge, helping teams uncover knowledge gaps, synthesize insights, and perform complex analysis faster.

Research mode enables deep, iterative exploration of complex topics beyond simple Q&A.

Enabling Research mode

1. Check if your Knowledge Agent has Research mode enabled (look for a Research toggle in the chat).
   1. If not, you'll need to enable research mode for the knowledge agent in the knowledge agent settings under Behavior.
2. Click the Research toggle to switch modes.
3. Type a detailed research question or topic in your next message. Guru will then craft a detailed research plan.
4. Review the plan and either:
   1. Click Confirm to start the research
   2. Click Revise to edit your original message and adjust the plan

✍️

Note

• You must have Owner access to a Knowledge Agent to enable the use of research mode for that agent.
• Do to the detailed nature of a research request, it can take up to 30 mins to complete. Research continues even if you navigate away from the chat.

What can Research do?

Complex Analysis & Synthesis:

• Comprehensive reports:"Research our competitor landscape and create a detailed comparison of their pricing models, target markets, and key differentiators"
• Multi-source validation: "Investigate all customer feedback from the past quarter and identify the top 3 product improvement opportunities"
• Policy analysis: "Research our security policies across all departments and summarize compliance requirements for new vendors"

Cross-functional Investigation:

• Process optimization:"Analyze our customer onboarding workflow from Sales handoff through Support resolution and identify bottlenecks"
• Troubleshooting complex issues:"Research all known causes and solutions for login failures across our different authentication methods"
• Strategic planning: "Investigate market trends, customer needs, and our current capabilities to recommend new product features"

📘

Best Practice

Guru attempts to create sections for the analysis based on the prompt provided. By using the word "section" you can inform research how to format our reports. For example, you can prompt it to create 3 "sections" and then specify what information you want in each section.

When to Use Research vs. Simple Q&A:

Simple Q&A: "What's our return policy?" or "Who's the account manager for Company X?" Research mode: Questions that start with "Analyze," "Investigate," "Compare," "Research," or "Create a comprehensive..." and require pulling together information from multiple sources to form a complete picture.

Frequently asked questions about Research

How do I know which sources a Knowledge Agent can access? Check the agent's configuration in your Admin settings, or ask the agent directly about its available sources.

Can I save or export chat conversations? Chat interactions are tracked for the individual on the left navigation bar of the Chat window in the web app. Questions that are identified will also be tracked in the Agent Center.

How does Guru Research support governance? Every chat and research interaction is logged in the AI Agent Center. Admins and experts can review conversations, retrain Knowledge Agents, and ensure high-consequence knowledge stays accurate and compliant.

Can I adjust the research plan before it runs? Yes. Guru shows the proposed steps first, so you can confirm or refine before execution.

Why did my research stop? Starting a new research plan with the same Knowledge Agent stops any previous research. Each agent can only run one research plan at a time.

What sources can Research Mode pull from? Guru's research can draw from any sources available to the Knowledge Agent you're conducting the research from. Sources such as Guru Collections, support tickets, uploaded files, and connected systems like Salesforce, Zendesk, Workday, or Confluence.

Does Research take into account the prompt for the relevant Knowledge Agent?

No, Research will not take into account the prompt of the relevant Knowledge Agent. It will however consider the tone of the prompt when generating.

Can I use research mode to analyze AI Agent Center data? Yes. You can use chat or research mode to conversationally query your AI Agent Center data for insights into Knowledge Agent performance. Ask natural language questions to review unanswered questions, analyze flagged responses, identify trending topics, track automation ROI, and filter by Knowledge Agent, time period, status, source, or asker. This helps you continuously optimize your Knowledge Agents based on actual usage patterns. Learn more in Guru's AI Agent Center.

Can we create a research-based knowledge agent that can access other knowledge agents without giving research access to everyone in the org? Yes. Create a research-specific knowledge agent and restrict its usage to approved researchers. That agent can perform research across other agents’ training information only when the person requesting the research has the appropriate permissions (e.g., Owner access on the target agent). This keeps research capability centralized while preventing permission bypass.