Fireside Chat, Interview
Bruno Gagliardo, Global Head of Data & AI @ Sanofi : Molecule to Market, Driving AI Transformation
- Strategic Vision: Sanofi CEO launched a five-year strategy to transform the company into a "pharma tech" leader powered by AI at scale, aiming to shorten the drug discovery lifecycle (currently 12–14 years) across research, development, clinical trials, manufacturing, and commercial operations.
- Scale and Scope: Sanofi operates with a $110 billion market cap, presence in 170 countries, and 85,000 employees, necessitating a massive migration from legacy on-premises systems to a cloud-based data mesh architecture.
- AI Readiness Framework: Bruno Gagliardo outlines three non-negotiable rules for AI potential:
- Context: Foundation models trained on trillions of tokens (equivalent to 12 million novels) require specific domain context to generate tangible insights.
- Interoperability: Adoption of standard formats like MCP to enable scale across diverse platforms.
- Trust and Safety: Security and quality cannot be sacrificed; the goal is to eliminate hallucinations while driving innovation.
- Cultural Transformation: Sanofi shifted to an agile, product-oriented operating model, prioritizing ruthless use-case prioritization and launching large-scale upskilling programs to boost data and AI literacy among 85,000 employees.
- Job Impact: Contrary to fears of displacement, the company found that AI augments human capabilities rather than replacing roles, with employees repositioning themselves to perform tasks faster using AI agents.
- Governance Strategy: To manage risks and the "shiny object" syndrome (e.g., the 2023 Gen AI hype), Sanofi established a strict governance mechanism with a minimum entry barrier of $100 million Business Impact Opportunity (BOI) for prioritized use cases.
- RAISE Framework: Sanofi implemented "Responsible AI at Sanofi for Everyone" (RAISE) before the EU AI Act, utilizing multidisciplinary boards (legal, cyber, architecture, data) to evaluate security, ethics, and maturity at every stage of the AI lifecycle.
- Production Barriers: A recent survey identified the top three inhibitors for moving GenAI to production as data quality/hallucination, security concerns, and cost management.
- Value Realization Examples:
- Regulatory Submissions: AI reduced the timeline for compiling clinical study dossiers for health authorities from 19 weeks to 5 weeks.
- Manufacturing: Agents assist in root cause analysis for batch failures and compile quality information for regulatory compliance.
- Research: Multi-agent systems sift through millions of datasets to generate insights, saving significant time and cost.
- Build, Buy, Borrow Strategy: Sanofi adopts a hybrid approach, building core blocks only when market maturity is low, and shifting to "buy" as partner solutions mature to ensure strategic control while leveraging external innovation.
- Agentic AI Deployment:
- Employee Concierge: An internal tool utilizing multiple backend agents to handle routine tasks like booking meetings, creating vacation requests, and managing IT tickets without leaving the interface.
- Research Partnerships: Strategic collaborations with partners like Okin to scale agentic capabilities for drug discovery.
- Risk Management: Guardrails for agentic systems include LLM-as-a-judge methodologies, prompt management standards, and automated drift detection to ensure models remain fit for purpose and ethical.
- Executive Alignment: The CEO and C-suite are actively engaged and data-literate, co-creating business cases with technology teams and serving on governance boards for high-impact projects.
- Business-IT Collaboration: Friction between business speed and IT security is mitigated through co-creation, agile product teams, and applying AI engineering (including "vibe coding") to internal IT processes.
- Financial Accountability: Value is tracked via governance bodies composed of C-suite members, with ROI reporting integrated into the standard operating rhythm for board-level scrutiny.
- Technology Ecosystem: Sanofi leverages a multi-vendor strategy including Snowflake, AWS, and specialized software partners to ensure scalability and rapid innovation without being siloed into a single provider.