The customer is the product team behind DRIFT, a digital regulatory intelligent filing tool developed in partnership with Slalom to support utilities in preparing complex regulatory filings. DRIFT is designed to streamline traditionally manual, resource-intensive processes such as research, drafting, review, and revision of filings submitted to utility commissions. The platform combines a modern web-based user experience with cloud-native backend services on AWS, and leverages generative AI to analyze past filings, regulatory guidance, and curated document collections. Due to confidentiality requirements, the end-customer organizations using DRIFT remain anonymous.

Regulatory filings for utilities are complex, iterative, and costly, often requiring extensive research, drafting, review, and revision across large volumes of historical filings and regulatory guidance. DRIFT’s product vision required a new Workspaces feature to support secure, multi-user collaboration while embedding generative AI directly into the drafting workflow. The challenge was to design and implement this capability without disrupting the existing Angular frontend or AWS-based backend, while also supporting both Slalom-hosted and client-hosted SaaS deployments. Additionally, the solution needed to evaluate and support multiple LLM options within Amazon Bedrock to balance performance and cost, all while adhering to AWS Well-Architected best practices and prioritizing security from the outset.

AWS provided the managed, scalable foundation required to deliver AI-assisted regulatory drafting at enterprise scale. Amazon Bedrock enabled secure access to Anthropic’s Claude large language models for text generation, summarization, and retrieval-augmented generation workflows, while allowing consideration of alternative models for performance and cost optimization. AWS-managed data and identity services supported secure, multi-user access and reliable backend operations. Building on AWS allowed DRIFT to remain fully cloud-native, align with existing architecture, and deploy consistently across multiple SaaS hosting models.

SnapSoft designed and implemented the Workspaces capability as a scalable, AI-enabled extension of the Drift platform. The solution integrated with the existing Angular frontend and aligned with the backend architecture using Amazon Aurora PostgreSQL and Amazon Bedrock. SnapSoft supported generative AI–driven drafting and research workflows, enabling users to analyze curated document collections, generate initial draft sections, and collaborate securely within shared workspaces. Identity and access management were implemented using Amazon Cognito, with per-user access toggling and feature-flag support to accommodate optional functionality. Throughout the engagement, SnapSoft collaborated closely with Slalom, providing architectural documentation, confirming model preferences, and supporting integration testing across environments.

• AI-assisted regulatory drafting: The Workspaces feature enabled generative AI–driven drafting and research workflows, reducing reliance on fully manual processes for regulatory filings.

• Seamless platform integration: The solution integrated cleanly with Drift’s existing Angular frontend and AWS-based backend, preserving architectural consistency.

• Secure multi-user collaboration: Built-in identity and access controls supported secure, role-based collaboration within shared workspaces.

• Flexible deployment support: The architecture supports both Slalom-hosted and client-hosted SaaS deployments, enabling broader adoption across customer environments.

• Future-Readiness: The AWS-native design positions Drift to expand AI-assisted capabilities across the full regulatory filing lifecycle, including deeper agentic workflows and end-to-end assisted drafting. By leveraging Amazon Bedrock and managed AWS services, the platform can adopt new foundation models, scale usage, and introduce additional features without significant re-architecture. This approach ensures Drift remains adaptable as regulatory requirements, customer needs, and generative AI technologies evolve.

• Amazon Bedrock
• Amazon Aurora PostgreSQL
• Amazon Cognito
• ECS Fargate
• Amazon DynamoDB
• AWS OpenSearch
• Amazon S3
• AWS CloudWatch
• AWS Lambda