Our client is an American multinational healthcare corporation offering a vast portfolio of products and services for the consumer, pharmaceutical, and medical devices markets.
The Challenge
Our client’s technology service (TS) organization works across 65 workgroups to create 10,000 to 12,000 statements of work (SOW) and proposal documents yearly. The process was manual and inefficient, with dispersed knowledge across multiple unconnected sources. To accelerate the process of designing proposals and solutions for upcoming business requirements, the client wanted to build a Generative AI-based TS CoPilot platform that would automate responses by tapping into past sources of knowledge, service, and application catalogs.
The Solution
The client partnered with Persistent to build the platform and streamline the overall process . The CoPilot platform taps into a variety of data sources including service catalogs, application and component inventories, and other knowledge sources to accelerate SOW and proposal creation. Using playbooks, the CoPilot leverages virtual agents powered by large language models (LLMs) to guide users in a step-by-step journey from business requirements to proposals.
The CoPilot platform also interacts with predictive models to estimate timelines, personnel requirements, and costs for proposed solutions. Using past solution diagrams and solution descriptions, the platform also generates this information for new business requirements, using multi-modal LLMs.
The Outcome
The TS CoPilot decreased human touch points by 70-80%, accelerating proposal response time by 90% and thereby speeding up the client’s ability to respond to SOW and proposal requests. The platform also connected disparate technology assets to act as a single source of truth, leading to higher data utilization.
Technology Used
- Azure Open AI GPT 4
- Qdrant Vector Db
- Langchain framework
- STreamlit UI
