InsightPulse

Inspiration

The pharma market moves faster than humans can track manually The pharmaceutical industry generates continuous signals-drug launches, label updates, regulatory approvals, pricing changes, and evolving communication strategies. However, these signals are: Spread across dozens of fragmented sources updated asynchronously and frequently, often noticed only after competitors act .Meanwhile, strategic decisions in pharma cannot wait for quarterly reports. This gap between market speed and human monitoring capacity inspired us to build an Always-On Market Intelligence Agent.

What it does

The Always-On Market Intelligence Agent continuously monitors, analyzes, and interprets what’s happening in the pharma market and converts it into actionable intelligence. In simple terms, it: Tracks brand updates, drug launches, and communication changes Detects changes compared to previous versions (what’s new, what’s different) Understands intent behind messaging (scientific, promotional, access-focused, patient-centric) Compares competitors and market trends against a selected pharma company Generates insights & recommendations, not just raw updates Shows sources transparently (URLs, timestamps) for trust and auditability

How we built it

The overall system is designed as a three-layer architecture. The first layer handles data ingestion, where information is collected from multiple external sources such as pharmaceutical company websites, social media platforms, LinkedIn posts, and other relevant public resources. This layer ensures that the system remains always up to date with the latest market activity. The collected data is then passed to the second layer, which forms the core intelligence of the system. A central orchestration agent distributes the data to multiple specialized agents responsible for analyzing different aspects of the market, including strategy changes, communication patterns, market trends, and product launches. The outputs generated by these individual agents are consolidated by an insights agent, which interprets and structures the findings. These insights are further processed by a deep analysis agent that compares the market intelligence with the company’s own product and strategy data stored in the database, ultimately generating actionable insights. The third layer represents the user interface, where outputs from all agents are visualized through an interactive dashboard. The dashboard presents key insights along with source URLs for transparency, highlights combined market trends such as targeted therapy areas and overall communication patterns.We also have a seperate deep agent attached to a chatbot to track activities of a particular comapny.

Challenges we ran into

market intelligence system faces three major challenges. First, noisy data: information ingested from websites, social media, and LinkedIn often contains promotional content, repetition, and low-signal updates, which can affect downstream agent analysis. Second, parsing heterogeneous data formats: inputs arrive as web pages, short posts, long articles, or documents, making it difficult to provide consistent, structured input to the main and specialized agents. Third, accuracy and quality of insights: combining outputs from multiple agents and aligning them with internal product data requires strong reasoning and context awareness; without this, insights may be shallow or misleading. Improving insight reliability is critical, as it directly impacts the value of actionable recommendations shown on the dashboard.