Answer components
Overview
The Answer in Connecty AI is where you see the final output from your natural language question. It turns your intent into a structured, transparent, and editable result.
Designed for both business users and technical analysts, it includes:
What the AI understood from your question
The logic behind the generated query
The data fetched from your warehouse
Tools to edit, audit, or reuse
The layout supports both quick validation and deep analysis.
Key capabilities at a glance:
Intent Graph:Visual breakdown of how the question was logically processed. (relevant in Deep Reasoning Mode).
Answer Summary: Clear business-focused explanation of the result.
SQL View: Generated SQL query with options to copy or edit.
Table View: Raw data in a structured, exportable table.
Components adapt to the reasoning mode (Simple or Deep) and chosen agent (Analyze or Discover).
1. Intent Graph
The Intent Graph shows how Connecty AI decomposes complex questions, especially in Deep Reasoning Mode. It maps the question into intents and sub-intents for full transparency.
Interactions
The Intent Graph isn’t just a visual summary, but it's also an interactive navigation tool. Users can explore the structure and reasoning behind their question by hovering over or clicking on nodes within the graph. Each node represents either the main intent or a sub-intent, and clicking them unlocks detailed insight into the reasoning path behind that node.
Hover
Main Intent node
The full natural language question (as interpreted by the AI)
Execution status (e.g., "Success", "Error", etc.)
Node type: marked clearly as INTENT
Sub-Intent node
The interpreted sub-question (e.g., derived measure or filtered metric)
Node type: labeled as SUB INTENT
Execution status (e.g., “Success”)
Click behavior
Main Intent
opens the Query Inspector, focused on the logic and execution related to the top-level question.
Sub-Intent - opens the Query Inspector scoped to that specific sub-intent, allowing the user to
Inspect the SQL query generated for that sub-step
View its reasoning flow, context, and grammar
Understand its role in the broader analysis
Best practice: Use the Intent Graph to explore complex queries rather than relying only on the table view.
For a full breakdown of Query Inspector functionality, refer to the dedicated article.
2. Answer Summary
The Response Description is the natural language explanation of the result returned by Connecty AI. It appears prominently at the top of the response view and summarizes the key insight derived from the query — essentially translating complex data logic into a concise, business-readable statement.
This feature helps both technical and non-technical users quickly understand what the response means, without needing to inspect the SQL or underlying data immediately.
Purpose
Bridge the gap between raw data and business understanding
Serve as a headline summary of the result
Communicate key dimensions, filters, and metrics used
Explain the response in terms that are aligned with business intent
Benefit
Description
Clarity
Helps non-technical users understand data results in plain language
Collaboration
Makes it easier to share results with business stakeholders
Decision support
Acts as a ready-to-use insight in reporting, dashboards, and presentations
Trust
Shows how the system interpreted the query and sets the stage for further exploration
3. Result SQL
The SQL View in Connecty AI provides a transparent look at the exact SQL query that was generated to answer the user’s question. Whether in Simple or Deep Reasoning Mode, this view allows users to validate how the system translated their natural language input into executable logic against the connected data warehouse.
This feature plays a central role in explainability, trust, and technical collaboration, bridging the gap between business intent and backend execution.
Benefit
Description
Transparency
See exactly what logic was executed to produce the result
Debugging
Identify potential mismatches between expected and actual logic
Validation
Confirm filters, joins, and aggregations match business understanding
Handoff
Share or reuse SQL in BI tools, notebooks, or downstream workflows
Learning
Understand how AI interprets natural language into data logic
3.1 Copy code
The “Copy Code” feature in the SQL View allows users to instantly copy the full SQL query generated by Connecty AI to their clipboard. This makes it easy to reuse, share, or debug the logic outside the platform — whether in BI tools, notebooks, email threads, or with data engineering teams.
What it copies
The full query currently shown in the SQL panel
Includes all filters, joins, groupings, and sort clauses
Any context-aware formatting or comments (if shown) will be preserved in the copied output
3.2 AI editor
📖 For a description about this tool’s functionality, refer to the dedicated AI Editor article.
3.3 Query Inspector
📖 For a description about this tool’s functionality, refer to the dedicated Query Inspector article.
4. Result Table
The Table View is where users can explore the raw data output generated by Connecty AI in response to their question. It presents the results in a structured, paginated table format, making it easy to scan, analyze, or export detailed results.
Purpose
Provide transparent access to the underlying dataset
Support manual exploration of results across multiple columns and rows
Enable spot-checking, data validation, or deep dives into individual records
Act as a launchpad for follow-up actions, such as refining questions or exporting data
Feature
Description
Column headers
Reflect selected dimensions and metrics, usually auto-labeled for clarity
Pagination
Navigable pages for datasets larger than the current view limit
Sorting
Columns may be sortable by value (ascending/descending)
Scrolling
Horizontal scrolling for wide tables
4.1 Export Data (coming soon)
The Export Data feature in the Table View allows users to download the results of their query for use outside of Connecty AI. Whether for reporting, deeper analysis, or documentation, this feature ensures users can take their data with them in a format that fits their workflow.
4.2 Result Table Statistics for Data Quality & Performance
Click the 'Statistics' icon on a result table to display metadata and data quality metrics:
Table-Level
Row count
In-memory size
Query execution time
Column-Level
Null count and %
Unique values
Min, max, avg (for numeric types)
This provides a quick overview of result characteristics without requiring manual inspection or additional SQL.
5. Result Validation & Assumptions (AI Inference)
This area appears below the result table and provides system-generated messages that highlight potential issues or clarify assumptions behind the data. e.g. logical derivation, temporal context resolution
Result validation
This block notifies the user when the returned results may have problems that could affect accuracy, privacy or other aspects of provided data. For example:
Unexpected data grain or aggregation – indicates if the result appears too detailed (e.g., row-level data instead of summary stats).
Sensitive fields warning – flags columns that might include personal identifiable information.
Review these warnings carefully before sharing or relying on the data.
Assumptions
This block clarifies how the system interpreted ambiguous queries or filled in missing logic. For example:
Filters applied – results only include discounted orders.
Calculation logic – revenue growth calculated as percent change from a previous period.
Confirm whether these assumptions align with your intent. If not, revise your prompt with AI Editor.
6. Answer Interaction Features (Call-to-actions)
Once you’ve reviewed the AI-generated answer—examining its logic, SQL query, metrics, visuals, and context—you can take additional actions through a set of interaction tools available in the top-right corner of each response. These tools support quality assurance, team collaboration, and iterative refinement.
Use these features to:
Mark a response as verified once it meets your expectations and delivers accurate results.
View Graph to open a visualization of all entities used in the answer.
Provide feedback if the answer is partially correct, misleading, or contains errors—helping the AI improve over time.
Copy a link to share the result with teammates or embed it into internal documentation.
Reply directly to refine the question, add filters, or ask for further breakdowns—keeping the conversation thread contextually intact.
6.1 Mark as Verified
When you are satisfied with a response, than you can mark it as verified. This signals to Connecty AI that the current query version aligns with the intended business logic and query style. Verified queries are flagged as trusted or final.
When clicked, this feature opens a verification screen that allows you to:
View all dimensions, measures, metrics, and relationships used to generate the answer.
Manually select the specific components you trust and wish to confirm.
This allows the system to:
Learn which individual metrics or dimensions are consistently correct
Safely reuse verified components in future answers
Support partial verification of multi-part responses
The selected items will be marked as "verified components", meaning they are now considered trustworthy building blocks for future queries across similar contexts.
Once you’ve selected the trusted components, click “Accept” to complete the process.
6.2 View Graph
Use this feature to open the Context Graph of this question
6.3 Provide Feedback
If the answer is incorrect, incomplete, or doesn’t reflect your expectations, the Provide Feedback feature helps Connecty AI improve. It’s intended only for reporting issues, not for compliments or general comments.
Only use feedback if something looks incorrect. Otherwise, try ‘Reply’ to refine your question.
This feature opens a structured feedback panel, where users can pinpoint what went wrong and optionally provide details:
Choose the option that best describes what went wrong.
By default, after submitting feedback, you are redirected to the AI Editor, where you can fix the query yourself. This ensures you can continue working while the issue is reported in parallel.
If you prefer not to open the editor, you can uncheck “Proceed to AI Editor after submitting”.
What happens to your feedback
Your report is collected securely and reviewed by the Connecty AI team.
It is used for debugging and improving model accuracy.
It does not affect your own data, nor does it immediately update the model.
6.4 Copy Link
The Copy Link feature allows users to generate a direct link to a specific answer, making it easy to share insights with teammates, document findings, or return to key results later.
When you click the 🔗 link icon next to an answer, a unique URL pointing to that specific message is copied to your clipboard. Anyone with access to the workspace and relevant data permissions can open the link and view the full query, visualizations, and breakdown.
6.5 Reply
📖 For a description about this feature, refer to Ask Question article.
7. Metricverse dashboard: New vs Reused entities
The Metricverse dashboard provides a detailed breakdown of all entities used to generate an answer. It is accessible at the bottom of every answer page and complements the “View Graph” option in the interaction toolbar.
Purpose
To show full lineage and transparency of the components used in generating the answer.
To help users audit whether metrics were newly created, reused from existing definitions, or already verified as trusted.
Key Features
Entity Categories are grouped for clarity.
Usage Labels: each entity is marked as New, Reused or Verified, so you always know the trust level of the components used.
Expand Groups: Clicking a group shows the list of all entities in that category.
Entity Details: Clicking an entity opens its full definition
Last updated