Building a Serverless, AI-Powered Case Management System on Google Cloud

Overview

A non-profit cultural heritage organization engaged with us for a Vertex AI Foundation project. The primary objective was to create a working model of Vertex AI’s main features to empower the organization to build and deploy a production-ready, AI-assisted case management system. The engagement was designed to be highly collaborative, with a focus on knowledge transfer to enable their staff to confidently develop additional capabilities independently. The project involved a series of deep-dive sessions and the delivery of a comprehensive assessment report.

The Challenge

The organization was facing significant challenges with their manual client intake, eligibility verification, document management, and financial disbursement workflows. These manual processes were not only time-consuming and inefficient but also prone to errors, leading to delays and a suboptimal experience for their applicants. They needed a unified and intelligent system to automate these processes, reduce the burden on their case managers, and provide a more streamlined and responsive service to the community they serve.

Our Solution

We designed a serverless, AI-assisted intake and case management system built entirely on the Google Cloud Platform. The solution was architected to be both intelligent and scalable, leveraging a suite of Google Cloud’s AI and application integration services. The core components of our solution included:

• Conversational AI Layer: We proposed using Dialogflow CX to create an intelligent and user-friendly interface for applicants, accessible via both voice and chat. This would handle the initial intake and screening process through natural, multi-turn conversations.
• Orchestration Layer: At the heart of the system, we designed a central “brain” using Cloud Run to execute business logic and connect the conversational AI with various backend services. This layer would automate data collection, eligibility validation by integrating with external APIs, and the initiation of financial disbursements.
• Data Management Layer: We designed a dual-database approach using Firestore for structured data, such as applicant profiles and case notes, and Cloud Storage for unstructured data, like uploaded documents. This ensures a centralized, real-time case database for case managers.
• Advanced Text & Sentiment Analysis: To enhance the safety and quality of interactions, we integrated the Google Natural Language API to perform sentiment analysis on applicant messages and automatically flag any abusive or concerning content for review.
• Security & Key Management: All API keys and credentials for third-party integrations were to be securely stored and managed using Google Secret Manager, ensuring a high level of security and compliance.

Impact

The project provided the organization with a clear and actionable blueprint for a modern, AI-powered case management system. The proposed serverless architecture was designed to be cost-effective and low-maintenance, allowing the organization to focus on their core mission rather than on managing infrastructure. By automating their key workflows, the solution would significantly reduce manual data entry, minimize processing times, and enhance the overall applicant experience. The centralized database and dashboard would provide their case managers with real-time access to case information, enabling them to work more efficiently and effectively.