CareLink Agent

Architecture
Overview and Problem Statement

Inspiration

Business Overview CareLink Health is a fictitious network of clinics and hospitals committed to delivering exceptional patient care. Recently, the company has been challenged by long call center wait times, fragmented appointment management, and difficulties handling urgent patient inquiries. CareLink Health needs a smarter, conversational solution that provides immediate, user-friendly interactions and streamlines patient support across its entire network.

Problem Statement Patients frequently experience delays when scheduling appointments, inquiring about prescriptions, or seeking emergency assistance. In addition, care teams are burdened by repetitive, administrative phone calls that disrupt care delivery.

Internal teams at CareLink Health encounter several hurdles in their day-to-day operations, including: Lack of Real-Time Visibility: Employees must manually retrieve and update data across multiple systems, which leads to delays and errors.

Inefficient Communication: Coordination between supply chain, logistics, and operations often involves multiple channels, resulting in miscommunications.
Reactive Workflows: Teams are forced to respond reactively to issues such as stock-outs or delayed deliveries instead of preventing them proactively. A conversational AI assistant that leverages Agentforce can streamline these tasks by enabling internal employees to access critical data, initiate workflows, and receive real-time updates—all through natural language conversations within Salesforce.

What it does

Provides agentic experience to patients where they can reach out to front desk 24x7 and get a response without being waiting for response. It helps stock inventory manager to keep track of essentials and place the order directly by asking agent to do so rather than logging into different system. Suppliers can stay up to date with incoming orders, it cuts down in the visits and dependency on inventory manager.

How we built it

We built three agents — one for Customers (Patients), one for Internal Users (Inventory Managers), and one for External Users (Suppliers) — across two Salesforce orgs. All agents were created in the first org (Hackathon Org), which holds the customer data for the first agent. Topics and actions built using flows fetch conversational context, and FAQs related to health insurance are served through a Data Cloud data library. The second org acts as a third-party system containing inventory data, exposed via APIs consumed by the second agent through API-based Apex actions. Suppliers interact with the third agent via Agentforce APIs. Deployment-wise, the first agent is live on Experience Cloud, the second is the default internal agent in Salesforce, and the third is deployed as an LWC in the second org.

Challenges we ran into

The assistant needs to work across internal websites and patient portals — each with different APIs, authentication methods, and UI limitations.
Flows like appointment booking, prescription inquiries must handle multi-turn conversations, interruptions, corrections and incomplete inputs.
API failures, timeouts, or backend issues, inconsistent responses from Agent needed more fine tuning the instructions.

Accomplishments that we're proud of

Able to build a high level framework for healthcare vendor which can provide a streamlined experience saving time and money. It was good to learn the usage of Agentforce APIs as opens up lot of possibilities of leveraging and connecting Salesforce with any system outside of Salesforce.

What we learned

Usage of Agentforce APIs and a potential usage of Data Cloud to support FAQ and keeping it dynamic.

What's next for CareLink Agent

If more time provided we would introduce a middleware to call external order management system and use Knowledge Articles along with Data Cloud to build a resource library for patients visiting the website and getting support. Introduce integration with more digital channels.