Skip to main content

Your first chat

The SixDegree chat interface lets you ask natural language questions about your infrastructure and get answers grounded in your knowledge graph. No query language required.

Before you start

You need a populated environment before chat is useful. Make sure you have:

  • Created an environment (see First Environment)
  • Connected at least one integration (Salesforce, Zendesk, Stripe, etc.)
  • Run discovery at least once so entities exist in the graph

Starting a chat​

Navigate to your environment at app.sixdegree.ai and click the Chat tab. Type your question and press Enter or click Send.

For terminal users:

degree chat

How it works​

When you send a message, the AI interprets your intent, queries your knowledge graph for relevant entities and relationships, and generates a natural language response. If you've configured MCP tools and the question implies an action, the AI can execute those tools before responding.

Example questions by role​

For revenue operations

  • "Which accounts renew in the next 30 days and have an open support escalation?"
  • "Which opportunities in negotiation have slipping product usage?"
  • "Show me the accounts with the largest subscriptions and no activity this quarter"

For support

  • "Trace this SLA breach to the incident behind it and the revenue it puts at risk"
  • "Which PagerDuty incident has tickets from more than one customer?"
  • "Show me the open tickets tied to our top-five accounts by ARR"

For operations

  • "Which vendors are we paying for that no team used this quarter?"
  • "Show me every contract renewing in the next 60 days and its owner"
  • "What's our spend by vendor across the finance integrations?"

For engineering

  • "What depends on the payments service, and who owns it?"

Chat features​

Contextual conversations​

The AI carries context across turns in a session. You can ask follow-up questions without restating what you're looking at:

You: Show me all accounts renewing this quarter

AI: I found 18 accounts with renewals in the next 90 days...

You: Which ones have an open support escalation?

AI: Of those 18, these 3 have an open ticket flagged as an escalation...

Entity references​

Click any entity name in the AI's response to open its full detail view in the graph.

Taking actions​

If you've configured MCP tools, the AI can execute actions on your behalf. The scope of what it can do depends on your autonomy settings:

Autonomy tierWhat the agent doesApproval required
ObserveReads and reports onlyNone
SuggestProposes actions with rationaleManual execution by you
ActExecutes non-destructive actionsNone
Act (destructive)Executes all actions including deletesConfirmation prompt

Example action prompts:

"Open a Zendesk ticket for the renewal risk on the Acme account"
"Draft a follow-up task on the opportunity stuck in negotiation"
"Create a GitHub issue in api-service about adding rate limiting"
Write better prompts

Be specific rather than vague. "Show me all accounts renewing this month" gets a better answer than "show me stuff". Use natural language as if talking to a colleague. "What's the status of the Acme renewal?" works well. Ask for explanations when you need context, such as "Explain the relationship between this account and its open tickets".

Troubleshooting​

SymptomWhat to check
AI can't find anythingConfirm discovery has run and you're in the correct environment. Trigger discovery manually from Environment Settings.
Responses seem inaccurateBe more specific or rephrase. Verify the discovered entities contain the data you're asking about.
MCP tools not workingCheck that endpoint URLs are correct and API tokens or credentials are valid.
Discovery is required

Chat answers are only as good as your discovery data. If you haven't run discovery recently, or if an integration was added after your last run, the AI may return incomplete or stale results.

Next steps​