Skip to main content

Starting Conversations

Getting the most out of the AI Assistant starts with knowing how to frame your questions. This guide covers how to open the chat panel, structure your questions, and use follow-up turns to drill into the details you need.

Opening the Chat Panel​

You can open the AI Assistant from any page in the Guardian Pro console:

  1. Click the chat icon in the bottom-right corner of any page.
  2. The chat panel slides open on the right side of the screen.
  3. If you have previous conversations, they appear in the conversation list.
  4. Click New Conversation or simply start typing to begin a new session.

The chat panel overlays the current page without navigating away, so you can reference the assistant's answers alongside the dashboards and data you are viewing.

Your First Message​

The best opening messages are specific and provide context about what you need. Here are some guidelines:

Be Direct​

The assistant works best with clear, focused questions:

Less EffectiveMore Effective
"Tell me about my AWS""What is my current health score and what is driving it?"
"Help me with costs""What are my top 3 cost reduction opportunities?"
"Security stuff""Do I have any S3 buckets with public access?"

Include Context When It Helps​

If your question is about a specific service, resource, or finding, mention it:

"Why is my RDS instance prod-database flagged as a security risk?"

"What would happen if the VPC vpc-0abc123 goes down?"

"How much am I spending on Lambda functions in eu-west-1?"

Ask One Thing at a Time​

The assistant handles focused questions better than multi-part requests. If you have several questions, ask them in separate messages:

Good: "What are my critical security findings?" ... then follow up with ... "How can I fix the most important one?"

Less effective: "What are my critical security findings, how do I fix them, and also what is my compliance score?"

Follow-Up Questions​

One of the most powerful features of the AI Assistant is multi-turn conversations. After receiving an initial answer, you can ask follow-up questions that build on the context:

Drilling Down​

Start broad, then narrow:

You: "What are my security risks?"

Assistant: Lists your top risks with severity levels

You: "Tell me more about the IAM root access key issue"

Assistant: Provides detailed explanation, affected resources, and remediation steps

You: "Can Guardian Pro fix this automatically?"

Assistant: Explains the remediation process and offers an action button

Changing Direction​

You can shift topics within a conversation. The assistant understands context switches:

You: "What is my compliance score for CIS?"

Assistant: Reports your CIS compliance score and failing controls

You: "Now tell me about my costs this month"

Assistant: Switches to cost analysis and reports spending data

Referring to Previous Answers​

Use natural references like "the first one", "that finding", or "the database you mentioned":

You: "What findings affect my RDS instances?"

Assistant: Lists three findings

You: "Explain the second one in more detail"

Assistant: Provides detailed explanation of the second finding

Understanding Response Indicators​

While the assistant processes your question, you see real-time indicators in the chat panel:

Tool Status​

When the assistant queries your environment data, a status label appears:

  • "Checking infrastructure health..." -- Retrieving your health score and contributing factors
  • "Analysing security findings..." -- Querying active findings and risk data
  • "Reviewing cost data..." -- Pulling spending trends and anomalies
  • "Checking compliance status..." -- Evaluating framework compliance scores
  • "Searching resources..." -- Querying your resource inventory

These indicators tell you the assistant is actively retrieving real data from your environment, not generating a generic response.

Streaming Response​

As the answer is generated, text streams into the chat bubble in real time. You can start reading immediately. The send button is disabled during streaming to prevent accidental interruptions.

Response Complete​

When the response is fully generated, you may see:

  • Action buttons -- Interactive buttons for navigation, remediation, or scanning.
  • Follow-up suggestions -- Recommended follow-up questions you can click to continue the conversation.

Tips for Effective Conversations​

Use Natural Language​

You do not need to use technical jargon or specific command syntax. Natural language works well:

"What is eating up my AWS bill?" works just as well as "Provide a cost analysis by service for the current month."

Mention Your Goal​

Telling the assistant what you are trying to achieve helps it provide more targeted guidance:

"I need to prepare for a SOC 2 audit. What compliance gaps should I address first?"

"We are trying to reduce our monthly AWS bill by 20%. Where should we start?"

"I want to make sure our production database is highly available. What should I check?"

Ask for Explanations​

If you are unfamiliar with a finding or recommendation, ask the assistant to explain:

"Why is it a problem that my security group allows inbound SSH from 0.0.0.0/0?"

"What does the 'AZ concentration risk' mean for my application?"

"Explain what an unattached EBS volume is and why it matters"

The assistant provides explanations tailored to the level of detail you request.

Ask for Next Steps​

After understanding an issue, ask the assistant what to do about it:

"What should I do to fix this?"

"Is there an automated remediation for this finding?"

"What is the recommended approach to resolve this?"

tip

The AI Assistant is designed for back-and-forth conversation. Do not try to ask the perfect question on the first try. Start with a broad question, review the answer, and then ask targeted follow-ups. This iterative approach produces the most useful results.

Starting from a Preloaded Context​

In some cases, the AI Assistant opens with a preloaded message based on where you were in the console. For example:

  • Clicking "Ask AI" next to a finding in the Action Centre preloads a question about that finding.
  • Clicking "Explain" on a risk in the Architecture Advisor preloads a request for a detailed risk explanation.

When a preloaded message appears, the assistant automatically processes it, so you get an immediate answer without typing anything.

Account Context​

The AI Assistant always operates in the context of the currently selected AWS account. If you switch accounts using the account selector:

  • The assistant reconnects with the new account's context.
  • A fresh conversation starts for the new account.
  • Previous conversations for the original account remain accessible in your conversation history.

This ensures the assistant never mixes data between accounts, even in follow-up questions.

Next Steps​