Capgemini-工业研发的数字化加速-如何快速、安全地从数据科学和人工智能项目中实现价值（英）-2023-18页_7mb

Summary of Digital Acceleration in Industrial R&D

Introduction
Digital R&D enables rapid innovation by speeding time-to-market and reducing costs, with data science and AI playing a pivotal role. Success hinges on strategic implementation, integrating tools across domains to deliver tangible value without compromising accuracy.

Prove Value Before You Commit

Analyze planned projects using Proof of Value exercises to assess data availability and potential usefulness. Prioritize projects that align with business goals and can deliver quick returns, such as through "Art of the Possible" workshops ensuring derisked investments and solid business cases.

Immediately Accessing the Right Data

Good data is foundational for any model. Data must be FAIR (findable, accessible, interoperable, reusable), of sufficient quality with trusted sources, supplemented by metadata for ease of use, and consistent across systems. This involves data screening, addressing privacy and security, and establishing data stewards to handle management.

The Right Type of Intelligence

Select models appropriate to the problem's context, data, and resources rather than pursuing the most powerful ones. Involve diverse data science experts, assess candidate algorithms quickly, and iterate early to ensure suitability and performance.

Deploying Models at Scale

Scale up model deployment by properly allocating compute resources, using frameworks like RAPIDE, and integrating models seamlessly into IT systems through methods like containerization. Ensure ongoing maintenance and support for robust functionality.

Overall Approach for Rapid Results

Accelerated digital R&D requires a portfolio strategy, cross-functional teams including strategists, data scientists, IT experts, and domain experts. Efficient skill allocation, as per the Pareto principle, frees domain experts to focus on core strengths while others handle data and model tasks, supported by governance frameworks for effective execution.