Your workspace and side panel share your private conversations, goals and processing history. Questions may concern the platform, a project or one selected object. Rename, pin, export, feedback and memory functions are available for integration, although not all appear in every browser view.
advanexus Intelligence
Turn a business question into a useful next step.
Ask about the selected data or task in your own words. Get explanations, SQL drafts and proposed next steps based on permitted context, while platform checks govern what can be executed.
The platform checks who is asking, their organization, project and row-level access. Intelligence receives the approved page context and selected-object references. It does not read arbitrary screen content, editor text or result grids.
Versioned guidance selects the relevant instructions and permitted tools for a request. Central rules use the configured model and define the scope of work. Different kinds of assistance do not create separate permissions or require users to choose a model.
The assistant can explain, draft or request an available tool. The platform checks the tool, its inputs, scope and rules again. Data previews and changes follow the confirmation or approval required for that particular action.
AI requests use the existing SQL, Dataset, Analytics, Assurance and Service Desk services. An explanation can be a complete answer, but a claim that something changed requires a completed action or tool result recorded by the responsible module.
For an analysis or dataset request, Intelligence uses the selected permitted source, schema and metadata. It explains what each row represents, the dimensions, metrics and chart suggestion, and checks the SQL for that database before display or preview. A draft is not proof that a query ran.
In asynchronous mode, the request is saved in the queue before processing begins. Progress is shown live or through periodic updates, while recorded events remain the source of truth. Stop, Regenerate, recovery of stalled work and protection against late results follow the same request history.
Continue the same question as your work progresses.
Your workspace and side panel share your private conversations, goals and processing history. Questions may concern the platform, a project or one selected object. Rename, pin, export, feedback and memory functions are available for integration, although not all appear in every browser view.
Start with the selected data and your current access.
The platform checks who is asking, their organization, project and row-level access. Intelligence receives the approved page context and selected-object references. It does not read arbitrary screen content, editor text or result grids.
Your identity, organization, active project and current access
The page you are using and the exact selected object
Limited source, SQL and metadata context with sensitive details removed
Tools that are available under the current configuration and rules
Match the assistance to the task, not a different set of permissions.
Versioned guidance selects the relevant instructions and permitted tools for a request. Central rules use the configured model and define the scope of work. Different kinds of assistance do not create separate permissions or require users to choose a model.
Review the proposal before an action changes anything.
The assistant can explain, draft or request an available tool. The platform checks the tool, its inputs, scope and rules again. Data previews and changes follow the confirmation or approval required for that particular action.
Goal → Grounded context → Proposal → Validation → Confirmation or approval → Action → Evidence
Use the same application controls, whether a person or AI starts the work.
AI requests use the existing SQL, Dataset, Analytics, Assurance and Service Desk services. An explanation can be a complete answer, but a claim that something changed requires a completed action or tool result recorded by the responsible module.
Sources and SQL Console: inspect metadata, check SQL, preview within approved limits and save a query.
Datasets, Data Quality and Analytics: find the exact resource, guide analysis and make only explicitly supported, controlled changes.
ANpy: check the structure of suggested Python code on the server; the assistant does not execute notebook cells.
Assurance and Service Desk: search within permitted scope and use supported case, evidence-package or ticket actions.
Get a SQL draft for the database you actually selected.
For an analysis or dataset request, Intelligence uses the selected permitted source, schema and metadata. It explains what each row represents, the dimensions, metrics and chart suggestion, and checks the SQL for that database before display or preview. A draft is not proof that a query ran.
Follow progress without losing the saved processing history.
In asynchronous mode, the request is saved in the queue before processing begins. Progress is shown live or through periodic updates, while recorded events remain the source of truth. Stop, Regenerate, recovery of stalled work and protection against late results follow the same request history.
Manage capacity and understand recorded usage.
Before contacting the model provider, the platform checks concurrent requests and request rates. Versioned policies can set limited or unlimited token capacity globally, per organization or per user. Usage and attributed costs are recorded; this is not a guaranteed spending ceiling or coordination across regions.
Keep assistance within the actions that are available today.
Intelligence does not search the web, operate a browser, read arbitrary screens or run an unrestricted autonomous workflow. It can use only available tools and never gains data or permissions that the current user lacks.
Drive, Sandbox and workflow preparation provide explanations only where no execution connection is implemented.
Starting, retrying or cancelling Pipelines and running Data Quality rules are not currently available AI actions.
Analytics changes are limited to an explicitly confirmed table-report draft for one exact dataset version, not general chart or dashboard creation.
Automatic retrieval from documentation, separate model-provider accounts per organization, fixed spending caps and a complete retention/legal-hold interface are not currently available.
Understand the suggestion and the evidence behind it.
An answer should use the selected source or object, distinguish facts from assumptions and provide checked SQL when requested. It explains what is being measured and why it is useful. Missing context must remain visible rather than being filled with invented data.
Shorten the path from question to action. Keep approval with people.
advanexus Intelligence explains permitted information and proposes an available next step.
The model does not approve its own changes. Platform checks validate the proposal, policies and authorized people govern changes, and the responsible module records what actually happened.
Business goal → Permitted context → AI proposal → Deterministic validation → Human authority → Registered action → Result and evidence
Next step
Start with the flow that cannot afford ambiguity.
Bring the systems, owners, rules, delivery obligations and evidence requirements that matter.
Discuss a critical flowIn brief
What does this part of advanexus do?
Ask about the selected data or task in your own words. Get explanations, SQL drafts and proposed next steps based on permitted context, while platform checks govern what can be executed.
What should I check before using it?
Confirmation, approval and evidence behavior depend on the exact registered action contract.