Contify's Ask Athena is purpose-built to deliver actionable Market & Competitive Intelligence (M&CI) for businesses, using a sophisticated multi-agent workflow that ensures maximum relevance, precision, and reliability in every client response. This article explains the workflow and illustrates how Ask Athena orchestrates its agentic flow, from processing the client’s query to generating a verified final answer.
📌 How Ask Athena operates through four specialized agents:
Agent 1: Categorization and Validation
Agent 2: Metadata Extraction and Database Retrieval
Agent 3: Web Enrichment [External Source Integration] (Agent 0)
Agent 3: Final Reasoned Response Generation
Step 1: Agent 1 [Categorization and Query Validation]
When a query is submitted, Agent 1 acts as the first decision-maker which uses a reasoning model. It receives the query along with the client’s company name, description, and industry specialization. The agent is responsible for:
Determining Scope: Categorizes whether the query is related to the client’s market and competitive landscape, general business-related, or irrelevant [outside the configured Contify scope].
Setting Guardrails: Ensures Athena only answers meaningful M&CI and business queries, informing users when their question is irrelevant, but providing an option to proceed.
Handling Ambiguity: Identifies missing contextual or time-sensitive information and prompts the client for clarification before moving forward.
This ensures only actionable and properly defined queries are processed further.
Step 2: Agent 2 [Metadata Extraction and Retrieval Preparation]
Relevant queries from Agent 1 move to Agent 2, which leverages advanced extraction techniques, again powered by a reasoning model:
Entity Extraction: Identifies companies, time frames, and client-defined category names [such as competitor buckets specified by the client].
Simplified Retrieval Query: Creates a stripped-down version of the query for efficient database matching, removing company names and dates.
Controlled Scope Filters: Applies filters using detected entities and specified or default time windows [usually last 12 months].
Company Category Context: Ensures generic queries like 'What are my competitors doing ?' correctly reference the client’s tracked entities by bucket name for granular, client-specific insight.
Agent 2 uses reranking [threshold ≥0.75] to keep only the most relevant data for answering queries, maximizing the quality of the resulting insights.
Step 3: Web Enrichment [External Data Integration] (Agent 0)
While contextually relevant platform stories/updates are used to generate a response for the query, Agent 0 fetches up-to-date, relevant information from the web via Large Language Model [LLM] with web browsing/search capabilities, enriching internal data with the latest developments [if any information is not available in the platform]:
Real-Time Search: Uses external search to pull timely news or business updates related to tracked companies and query buckets.
Query Optimization: Replaces placeholders or bucket names with their actual tracked values for targeted, precise searches.
Synthesizing Insights: Combines external findings with internal database results.
This merging of in-house and external data ensures Ask Athena’s responses are timely and comprehensive.
Step 4: Agent 3 [Response Generation with Source Attribution]
Once all inputs are prepared, Agent 3 synthesizes the selected data using a reasoning model to generate the final response:
Structured Output: Presents information in predefined formats such as summaries, lists, comparisons, based on the query type.
Source Mapping: Each insight is attributed to its source, combating hallucination and building trust with clients.
Business Relevance: Answers are always aligned with client context, business needs, and industry focus.
Agent 3 orchestrates the final output, balancing clarity, substance, and transparency for robust client decision-making.
Optimized for Speed: Parallel Processing
Ask Athena leverages parallel execution, where Agent 1 and Agent 2 synchronize the query and metadata, while Agent 0 runs concurrently, resulting in full workflow completion within 30–35 seconds. (post-processing by Agent 3)
Why Ask Athena AI Agentic Flow Matters
Uncompromising Relevance: Queries are contextualized to industry and business focus.
Timely Intelligence: Up-to-the-minute answers through dynamic web enrichment.
Trustworthy Answers: Each point is verified by its data source, eliminating hallucinated insights.
Proactive Query Guidance: Agents flag ambiguous or incomplete queries, delivering reliable answers.
This multi-agent system ensures Ask Athena delivers business intelligence that is:
Precise
Fast
Dependable
Custom-fit to every client’s needs