User Access Information 
Specifies what data each user has access to based on SSO.
SUPERVISOR AGENT
Structured Data Analytics Supervisor Agent Live
Manages the end to end lifecycle of user questions on structured data, ensuring accuracy, efficiency, and continuity across interactions.
The Structured Data Analytics Supervisor Agent orchestrates natural language questions on structured datasets by applying guardrails, classifying use cases, filtering irrelevant requests, and checking cached responses before triggering workflows like Text2SQL or visualization. Configured with metadata, business rules, compliance policies, and role-based access, it uses required inputs (queries) and optional context to deliver accurate, efficient, and consistent multi-step analytics.
Super Agent Actions
Preprocessing
Validates user questions and access, applies compliance guardrails, and ensures sufficient information is available to proceed.
Generate a Plan
Classifies the question to the correct business use case, checks for cached results, and identifies the appropriate downstream workflows (e.g., Text2SQL, Visualization).
Execute a Plan
Orchestrates the selected workflows, providing downstream agents (e.g. Instance of Text2SQL agent, Visualization agent) with the required metadata, use case details, and user profile.
Synthesize a Response
Aggregates outputs from downstream agents, preserves session context, and delivers a coherent, accurate, and consistent response to the user.
Configurations
The configurations that allow this agent to be used for a specific business context and data source
User Access Information
Use Case Details
Details of each use case for appropriate routing including what is the exact scope of the use case and how the use case can be triggered.
Guardrails
Different compliance, security, and content restrictions set at the AbbVie and use case level.
Summarizer templates
Predefined formats for each use case defining the summarization style for the data summary.
Metadata: Enhances AI's understanding by supplying instructions, business term definitions, synonyms, acronyms, and links between data and business terms using business rules. This information must be provided by a business SME.
New Configuration Instructions
Configurations for this agent are determined by dataset or use case. To configure this supervisor agent for a new dataset or use case, reach out to the LAILA team with the use case details and discuss the additonal information needed to reliably orchestrate the users' questions.
Configured Data Sources
Oncology Subnational Sales FFP
Prompt Guide
Required Inputs
User Question
Description
The query or question provided by the end-user in plain, human-readable language
Accepted Values
String containing the user’s input question
User Access Information
Description
Metadata related to the user’s identity and access privileges used for authorization
Accepted Values
User email
Optional Inputs
Context from Previous Question
Description
Details retained from the user’s earlier query to maintain continuity in multi-step conversations
Accepted Values
User question, parameters, or structured data carried over from the previous question
Use Case Specific Parameters
Description
Domain-specific inputs that refine or filter the query.
Accepted Values
TA (Therapeutic Area): Oncology Franchise: Franchise V Indication: AML, CLL, or All
Example Prompts
Hello!
What can you do?
How many HCPs wrote Venclexta in 2025?
Related Information
Tools Used The smaller, reusable components that this agent can utilize to perform repetitive tasks without using GenAI.
Guardrails
Applies rules to validate, restrict, and ensure safe, compliant user queries before execution
Cache Utility
Stores and retrieves past query results to reduce latency and improve response time.
Error Logger
Captures, stores, and reports agent errors for debugging, monitoring, and compliance tracking.
Response Formatter
Structures agent outputs into Dashboard accepted format
Summarizer
Condenses lengthy outputs into concise, user-friendly summaries for quick interpretation
Recommendation
Fetch recommended question using similarity search based on the question which is asked by user
