Schneider Electric automates Salesforce account hierarchy management with generative artificial intelligence (AI) using Amazon Aurora and Amazon Bedrock

Schneider Electric, a leader in energy management and industrial automation, sought to improve their CRM system by accurately linking customer accounts to their parent entities. This process, crucial for optimizing sales operations and resource allocation, was manually intensive. To address this, Schneider Electric leveraged generative AI and large language models (LLMs) in April 2023. Initially, they used Flan T5 LLM in Amazon SageMaker Jumpstart, proving the project's feasibility. Subsequently, they enhanced their solution using Amazon Aurora and Amazon Bedrock integrated with Salesforce CRM. Amazon Bedrock, a managed service providing high-performing foundation models, facilitated building secure and responsible AI applications. Amazon Aurora Serverless V2, a relational database, supported storing machine learning (ML) model embeddings using the pgvector extension for efficient similarity searches. Schneider Electric used pgvector to store embeddings from LangChain, enabling effective customer account hierarchy mapping. The solution involved AWS Batch jobs processing Salesforce account information, making recommendations using Anthropic Claude 3 on Amazon Bedrock, and storing results in Aurora and Amazon S3. A custom Streamlit application displayed these recommendations for user review. This automated process reduced the time to update account hierarchies from seven to three minutes per account. The LLMs played a critical role in generating accurate account hierarchies. By integrating search engine results, Dun & Bradstreet API data, and vector searches in Aurora, the system inferred correct parent entities. The flexibility of Amazon Bedrock allowed Schneider Electric to switch between different LLM models to optimize cost and performance. This approach led to a 60% cost reduction and improved efficiency in account hierarchy management. Overall, the project highlighted the benefits of using AWS services to innovate and streamline operations. Aurora Serverless v2 with pgvector and Anthropic Claude 3 Sonnet on Amazon Bedrock provided a scalable, cost-effective solution for managing customer account hierarchies. The integration of advanced AI models and serverless architecture minimized operational overhead, enhanced data accuracy, and ensured the system's agility and scalability.