This case study examines the implementation of GenAI across three major enterprises - AstraZeneca (pharmaceuticals), Adobe (technology), and Allianz Technology (financial services) - providing valuable insights into how different industries approach LLMOps and AI deployment at scale. ## Overview of Company Approaches ### AstraZeneca's Research Assistant Success AstraZeneca's approach focused on solving specific research and development challenges in the pharmaceutical industry. Their primary success story revolves around a research assistant tool that helps scientists with literature searches and data analysis. The implementation started with a focused pilot of 40 users but quickly went viral within the organization due to its effectiveness in saving researchers weeks or months of time in literature searches. Key aspects of their approach included: * Early engagement with senior leadership through AI bootcamps and prompt engineering training * Careful prioritization of use cases based on real scientific needs * Early legal engagement enabling rapid scaling * Silent launch that quickly grew to thousands of users * Iterative development adding new features and reasoning capabilities ### Adobe's Legal-First Framework Adobe took a unique approach by placing legal considerations at the forefront of their AI implementation strategy. Their first major model, Firefly for image generation, was developed with a strong focus on legal compliance and ethical considerations. They made the strategic decision to only train on licensed content, which enabled them to: * Provide commercial indemnification for their AI products * Build trust with the creative community * Develop a clear framework for evaluating AI use cases Adobe developed an A-to-F framework for evaluating AI implementations: * A: Team wanting to use the technology * B: Specific technology being used * C: Input data sources and types * D: Expected output data * E: Target audience * F: Final objective ### Allianz Technology's Cultural Transformation Allianz approached AI implementation as a cultural transformation rather than just a technology rollout. Their experience with Amazon Q for code modernization highlighted several key principles: * Treating AI adoption as a fundamental change in work practices rather than just a tool deployment * Focus on removing "nay tasks" (tedious work) to enable more meaningful work * Clear communication that the future belongs to developers who embrace AI tools * Implementation of "Gen AI hacks" sharing sessions to spread knowledge and excitement ## Key LLMOps Implementation Lessons ### Model Selection and Architecture The case study reveals sophisticated approaches to model selection across all three companies. Key considerations included: * Model agnostic architecture enabling easy swapping of models * Balance between latency, cost, and accuracy requirements * Recognition that model choices will evolve rapidly * Need for flexible architectural models that allow component replacement ### Scaling Strategies Successful scaling required careful attention to: * Early identification of potential blockers * Platform-level approvals rather than individual solution approvals * Global considerations for different regulatory environments * Balance between speed and standardization * Focus on smaller, impactful use cases rather than just large flagship projects ### Legal and Compliance Integration A critical success factor was the early and deep integration of legal teams: * Legal treated as enablers rather than blockers * Early engagement to accelerate implementation * Understanding legal teams as both stakeholders and users of AI systems * Platform-level legal approvals to enable faster scaling ### Cultural and Organizational Aspects The companies emphasized the importance of cultural transformation: * Moving from "digital by default" to "AI by default" * Focus on democratizing AI usage across the organization * Building organization-wide AI capabilities rather than limiting to specialist teams * Creating mechanisms for sharing success stories and best practices ### Measurement and Success Criteria Success evaluation included: * Clear prioritization of use cases based on value and scalability * Quick decision-making on whether to continue or stop experiments * Focus on measurable impact (time saved, quality improvements) * Balance between quick wins and strategic long-term goals ## Challenges and Lessons Learned Key challenges identified included: * Need for faster implementation cycles rather than traditional IT project timelines * Importance of data organization and accessibility * Balance between standardization and innovation * Managing the rapid evolution of AI technologies * Global regulatory and compliance variations The case study emphasizes the importance of: * Starting with clear problem definition rather than technology-first approaches * Building flexible architectures that can evolve with technology * Creating strong cross-functional partnerships * Maintaining focus on business impact and value creation * Treating AI implementation as a transformational journey rather than a technology project This multi-company case study provides valuable insights into how different industries approach LLMOps, highlighting both common challenges and industry-specific considerations in implementing AI at scale.