Conversation

Interface Layout

The Conversation Home is your central hub for starting, viewing, and collaborating on conversations.

Left Sidebar — Starter Pane

• Recommended: AI-generated questions tailored to your role and usage. Updated daily.

• Starred: Your pinned conversations for quick access.

• Recently Used: The last 10 conversations for fast resume.

Center Panel — Chat Window

• Large, responsive input for natural-language queries

• Full conversation thread with preserved context

• Rich responses (tables, charts, code blocks)

• Contextual action buttons and smart suggestions

• Real-time agent activity (“thinking”) visibility

Top Navigation Bar

• Logo: Return to Conversation Home / start a new conversation

• Profile & Settings: Manage preferences

• Notifications: Shared items and system updates

• Help: Documentation and user guides

Multi-User Conversation Features

• Invite users to any conversation (live or async)

• Real-time collaboration with teammates

• @adm mentions to invoke AI in team threads

• Seamless human + AI interaction in one place

Starting and Managing Conversations

Start a New Conversation

• From Home: Click the logo or New Conversation, type your query, then Send.

• From History: Open History in the sidebar > + New Conversation > type and Send.

• Using Slash Commands: Type / to open the command palette, pick a command, follow prompts.

Manage Conversations

• Rename: Open a conversation > Click the title > Enter a descriptive name > Enter.

• Delete: In History, hover over a conversation > click 🗑️ > confirm.

Important: Deletion is permanent (messages and context are lost).

• Star: In the header, click ⭐ to pin/unpin. Starred items appear in the Starter Pane.

Conversation Titles

Auto-generated titles summarize content. You can rename anytime to match team conventions.

Response Styles

Response styles tailor tone and depth to your audience or task.

Available Styles

1. Data Engineer: Technical depth, code/SQL where relevant, performance tips, lineage/schema details, troubleshooting.

2. Business Analyst: Business-friendly language, summary-first, visuals (charts/trends), executive-ready recommendations.

3. Data Steward: Governance focus, compliance terms, policy best practices, metadata and stewardship guidance.

4. Casual: Conversational, concise, low-jargon, ideal for quick lookups.

Select a Style

Use the Response Style dropdown in the conversation header. The choice applies to the current conversation and persists until changed.

Thought Process Visibility

The Agent Thinking panel explains how ADM interprets and answers your query.

Benefits

• Understand the reasoning path

• Verify the agents, tools, and data sources used

• Debug unexpected results

• Improve prompts by seeing how questions are parsed

• Support audit and governance reviews

What You’ll See

• Question analysis and time frame interpretation

• Agent selection and rationale

• Execution plan and data retrieval steps

• Processing/transformations

• Response formatting plan

Interact with the Panel

• Click the header to expand/collapse

• Review during debugging, audits, or training

Example Query: “Show me failed data quality policies from the last 24 hours.” Agent Thinking: Identify policy results → filter status=failed and time=24h → query Data Quality Agent → prepare table view.

Sharing and Collaboration

Share a Conversation

1. Open the conversation → click Share.

2. Search and select recipients.

3. Set permissions:

   • View Only: Read-only

   • Can Comment: Add comments only

   • Can Edit: Read, continue, and interact with ADM

4. (Optional) Add a message → Share.

Manage Access

• View who has access and their permissions

• Modify permissions or remove access

• Transfer ownership when needed

Collaborative Features

• Real-time updates without refresh

• @mentions to notify teammates

• Threaded comments for complex topics

• Attribution for actions and messages

• @adm to bring AI into team discussions

Note Sharing requires proper view:user permissions. Contact your administrator if access is restricted.

Export and Documentation

Download as PDF

Export complete conversation history with preserved formatting.

Includes

• Full message history with timestamps and authors

• Tables, charts, and code blocks

• (Optional) Agent Thinking steps

• Suggested actions and follow-ups

How to Export Open the conversation > Download as PDF > file downloads (e.g., ADM_Conversation_2025-10-25.pdf).

Use Cases

• Compliance/audit trails

• Stakeholder updates and summaries

• Knowledge base and training material

• Report appendices and documentation

Dashboard View (Experimental)

Switch to an interactive dashboard summarizing conversation analytics (counts, response times, agents used, asset distributions, timelines). Export as image for presentations and reports.

Features may change; share feedback to guide improvements.

Conversation Actions and Follow-ups

Contextual Actions

ADM suggests actions based on the current topic:

• Asset management: Tag assets, create workflows, visualize pipeline/lineage

• Policy management: Create/modify policies, schedule executions

• Investigation: Open asset details, inspect lineage, view incident reports

Suggested Follow-up Questions

After each response, ADM proposes 3–5 targeted follow-ups:

• Detail-oriented: e.g., “List all columns and types for CUSTOMER_ADDRESS.”

• Action-oriented: e.g., “Create a data quality policy for this table.”

• Exploratory: e.g., “Which other tables reference customer addresses?”

Click any suggestion to add it to the conversation.

Response Feedback

• 👍/👎 to rate quality

• Copy to clipboard

• Regenerate to try a new answer (Feedback informs future improvements.)

Knowledge Base Integration

Add Documents

Support formats: PDF, DOCX, TXT, MD Settings > Knowledge Base > Upload Document > wait for indexing.

What Indexing Does

• Extracts text

• Builds searchable embeddings

• Surfaces key concepts and relationships

Use in Conversations

ADM automatically references relevant documents after indexing and may cite them in responses. Verify by asking about uploaded content and checking the Agent Thinking panel.

Best Practices

Write Effective Questions

• Be specific: “Show failed DQ policies for customer tables in the last 7 days.”

• Provide context: “Most accessed Snowflake tables in the sales database?”

• Use natural language: Ask as you would a colleague.

Organize Conversations

• Rename clearly: e.g., “Q4 2025 Sales DQ Analysis,” “Customer Address Standardization”

• Star key threads (references, live investigations, major findings)

• Archive/Delete outdated or duplicate threads

Collaborate Well

• Share for multi-party investigations, onboarding, cross-functional work

• Set permissions appropriately (View / Comment / Edit)

• Add context and @mention teammates

Optimize Performance

• Start new threads for very long or unrelated topics

• Break complex prompts into smaller steps

• Constrain time ranges and sources where possible

Troubleshooting

ADM not responding / timing out

• Simplify the query or split steps

• Check your network connection

• Try again shortly (system load)

• Start a new conversation if the thread is very long

Inaccurate responses

• Be more specific and add context

• Confirm required data sources are connected

• Ensure relevant agents are enabled

• Review Agent Thinking to verify reasoning and sources

Cannot share conversations

• Confirm view:user permission

• Verify the recipient exists in your org

• Re-authenticate and retry

• Contact your administrator if the issue persists

PDF export issues

• Allow pop-ups / check download permissions

• Ensure local disk space

• Try another browser

• For very long threads, export sections

Dashboard view problems

• Refresh or clear cache

• Try a different browser

• Remember it’s experimental and may have limitations

Getting help

• Use Help in the top bar for docs

• Review suggested follow-ups and Agent Thinking

• Contact your admin for permission issues

• Provide conversation IDs when reporting problems to support