Norstella has launched Atlas, an agentic AI platform that converts 30 years of proprietary drug development data into finished analysis. The system processes natural language queries and returns sourced, actionable outputs for competitive intelligence and strategy teams, replacing manual workflows that once took analysts days or weeks.
Kris Kaneta, the company's chief product and innovation officer, and Dan Chancellor, VP of thought leadership, described Atlas as "the agentic expression of this highly valuable and unique data foundation." The platform draws on Norstella's existing data network spanning clinical development, regulatory, commercial, and market access, linked at the entity level across the full drug lifecycle.
How Atlas processes drug development data
Users interact with Atlas through plain language questions. The system returns detailed, sourced responses within minutes. "Picture working alongside an extremely capable and indefatigable colleague," Kaneta and Chancellor said. "You ask a question in plain language, and within minutes you are presented with a detailed response."
Atlas joins a growing category of AI agents and automation platforms that go beyond simple retrieval to execute multi-step reasoning tasks. The company deliberately built the system on proprietary data rather than public corpora. "Most AI tools in this space are essentially wrappers," Kaneta and Chancellor said. "Atlas can't be replicated by pointing a model at public sources."
The Clarity Test: keeping outputs verifiable
The team learned early that speed alone does not earn trust with high-stakes users. "When we started building Atlas, the focus was on capability: how much data an agent could reason across, how fast it could produce a finished output," they said. "What we learned quickly is that capability alone doesn't earn trust."
That lesson produced the Clarity Test, an internal benchmark every agent must pass before shipping. Every Atlas output must meet four criteria: cited (traceable to a source), consistent (grounded in a stable context layer), contextual (matched to the actual decision workflow), and consequential (actionable, not just confident-sounding). As the system scales, the company has added agents that monitor other agents to maintain quality. The platform applies AI for healthcare decision-making where data integrity directly affects patient outcomes.
First agent live, more in development
The first deployed agent, Atlas CI, targets competitive intelligence teams. It generates drug profiles, catalyst timelines, and executive briefings that previously required days of manual work. "This allows CI teams to rebalance their efforts towards the implications of their research," Kaneta and Chancellor said. Future agents will focus on business development and licensing, feasibility, portfolio strategy, and protocol design.
Real-world data integration is planned for later versions, adding patient journey insights and unmet need analysis. The company expects the platform to improve the speed and quality of drugs moving through development, though it declined to give a five-to-ten-year outlook, noting the pace of change in pharma AI.
Why this matters for IT and development professionals
Atlas demonstrates a practical architecture for enterprise agentic AI: ground outputs in proprietary data, enforce verifiability through a hard benchmark, and build oversight agents that scale with the system. For teams building AI tools in regulated industries, the lesson is direct - capability without auditability will not earn adoption. The platform's design, with agents that monitor agents and a stable context layer instead of a shifting public corpus, offers a reference pattern for deploying AI where errors carry real costs.
Your membership also unlocks:
