Scaling AI across the enterprise (an adoption model)
The problem
Drug development takes ~10–15 years and ~$2.6bn, with ~90% of candidates failing — and scientific data sits in disconnected silos that block AI from adding value.
The AI approach
Rather than one tool, AstraZeneca built the foundations: an enterprise data-and-AI architecture and a proprietary Biological Insights Knowledge Graph (BIKG) to generate novel target hypotheses, plus partnerships (e.g. BenevolentAI) and firm-wide generative-AI upskilling.
Evidence it works
The BenevolentAI collaboration yielded a validated AI-generated target for chronic kidney disease that entered AstraZeneca’s portfolio; by mid-2025 the firm reported upskilling ~12,000 employees on generative AI with 85–93% reporting productivity gains, and piloting AI assistants for tasks like 3D CT-scan analysis. Notably, it ran ~a year of proofs-of-concept to identify high-impact use cases first.
What “good” looks like
A clean data foundation, a portfolio of validated high-value use cases, and a workforce equipped to use AI — adoption measured by real outcomes, not tool counts.
Feasibility & cost shape
A multi-year, top-down programme; the “plumbing” (data architecture, governance, skills) is the real investment.
The most useful example here isn’t a single capability — it’s the adoption model: fix the data foundation, diagnose high-value use cases first, and upskill people. That’s precisely our enablement thesis.
Based on publicly reported information about the AstraZeneca work.
This is an industry example included for illustration. It is not a Leia Intelligence project, and no client of ours is implied. Figures are as publicly reported by the original parties.
Sources: AstraZeneca (data science & AI) · Emerj · Klover.ai (BIKG analysis) · clinical research trade press (upskilling figures)