CarePulse Copilot

Inspiration

The gap between clinical visits is often where patients feel the most uncertain, isolated, and overlooked. We noticed that care coordinators and nurses are overwhelmed, often juggling dozens of patients without the tools to personalize outreach at scale. With rising chronic conditions, post-discharge complexities, and the demand for value-based care, we wanted to create something that bridges the communication gap — while reducing the administrative burden on care teams. That’s where CarePulse Copilot was born.

What it does

CarePulse Copilot is an AI-powered assistant for care coordinators, built on Salesforce Health Cloud and Agentforce. It: Surfaces key patient context from EHR, lab results, social risk data, and care plans before the call. Guides care coordinators in real time during outreach, detecting patient sentiment and health signals like medication adherence or symptom worsening. Suggests empathetic, medically appropriate responses. Automatically updates patient records and sends tailored education or follow-up links post-call. Personalizes each interaction so patients feel truly supported, not just checked in on.

How we built it

Platform: Salesforce Health Cloud + Agentforce as our core foundation. Data Integration: We pulled in EHR, care plans, and social determinants data via FHIR and Salesforce APIs. Real-Time AI Assistant: Leveraged a large language model (LLM) fine-tuned on clinical guidelines and nurse-call scripts. Live Sentiment Detection: Used audio and text-based cues to analyze emotional tone and conversation intent. Smart Follow-ups: Automated resource sharing, task creation, and call summaries into Salesforce records. The frontend uses a sleek Agentforce UI for seamless integration into daily workflows.

Challenges we ran into

Structuring clinical data in a way the LLM could safely understand without hallucinating. Ensuring HIPAA compliance while handling real-time patient context. Balancing automation with human empathy — making sure Copilot felt like an assistant, not a replacement. Training the AI to understand nuanced healthcare language and patient sentiment in varied tones and dialects.

Accomplishments that we're proud of

Delivered a working Copilot that seamlessly augments care coordination workflows. Achieved contextual awareness using real-time health data, without manual prep from staff. Built a dynamic prompt system that keeps conversations medically accurate and empathetic. Created a tool that reduces burnout for care teams while improving patient engagement.

What we learned

Human-AI collaboration works best when AI plays the role of a guide, not a decision-maker. Care coordinators love technology that feels like it has their back — especially during emotionally charged calls. Real-time AI in healthcare is possible — with the right boundaries and data governance. Integration into existing workflows (like Salesforce Health Cloud) is critical for adoption.

What's next for CarePulse Copilot

Expand to multilingual support to assist non-English speaking patients. Integrate voice-based Copilot features using NLP for phone calls, not just chats. Add predictive triage scoring to help coordinators prioritize outreach. Partner with providers and payers to test real-world impact on readmission rates and care quality metrics. Open up Copilot APIs for integration into other patient engagement tools or CRMs.

Built With

• agent