CanAssist

Inspiration

The inspiration for this project stemmed from the growing need to simplify how constituents access and interact with government services. With services spread across federal, provincial, and municipal levels—each with its own systems, processes, and points of contact—users often face confusion, duplication, and delays. We envisioned a solution that mirrors the convenience of modern digital platforms but tailored for public services: a unified, intelligent assistant that could guide users end-to-end, no matter the level of government or complexity of their need.

Advancements in conversational AI, agent orchestration, and Salesforce’s Agentforce framework made it feasible to reimagine this experience—not only for constituents but also for public servants, enabling faster collaboration and better case resolution across government silos. The goal was to bridge the gap between government complexity and citizen expectations using technology that feels personal, seamless, and proactive.

What it does

The solution provides a single, intelligent point of contact for constituents to access government services across all levels—federal, provincial, and municipal. It uses multiple specialized Agentforce Agents, each designed to handle specific service areas. For the purpose of this Hackathon, the Agents are all hosted within one Salesforce instance but in a real-world scenario, they would be working together from separate instances. These Agents draw from Knowledge and S-Object data to deliver accurate, contextual responses. An Orchestrator Agent manages coordination by invoking AI Agent Actions in Flows, seamlessly routing tasks to the appropriate Agents and consolidating responses—giving users a unified, streamlined experience.

How we built it

We built the solution using Salesforce Agentforce, designing multiple specialized Agents with distinct skills and goals—all hosted within a single Salesforce org. These Agents leverage Knowledge Articles and S-Object data to provide contextual, accurate responses about services across federal, provincial, and municipal levels.

At the core was an Orchestrator Agent, which used AI Agent Actions embedded in Salesforce Flows to coordinate with other Agents behind the scenes and deliver unified responses to the user. This setup allowed us to simulate a seamless, one-agent experience while actually distributing tasks across specialized Agents. The architecture focused on modularity, allowing for future scalability into a multi-org Agentforce network integrated via APIs and orchestrated by MuleSoft.

Challenges we ran into

A key challenge was orchestrating agent-to-agent communication across different layers of government while maintaining context and delivering a seamless experience for the user. Intra-org orchestration using AI Agent Actions in Flows was straightforward, but cross-org communication introduced complexity—particularly around authentication, session handling, and preserving context between Agents.

Although full API orchestration is not part of the official demo scope due to platform rules, we still built a working proof of concept using the Agentforce API. This allowed us to validate the technical feasibility of enabling Agents from separate Salesforce orgs to collaborate in real-time, proving that such an architecture can scale to support complex, distributed government service delivery.

Accomplishments that we're proud of

We’re particularly proud of the seamless integration and orchestration we achieved between multiple Agentforce Agents within a single Salesforce instance. This allowed us to create a unified, intelligent experience for users seeking government services at all levels. By leveraging AI Agent Actions in Flows, we were able to automate agent-to-agent interactions efficiently, delivering accurate, context-aware responses to constituents.

Additionally, despite API orchestration not being part of the official demo scope, we successfully built a proof of concept using the Agentforce API to enable communication between Agents across different Salesforce orgs. This breakthrough not only validated the scalability and potential for future cross-org integrations but also opened the door to future interoperability with other AI Agents, such as Gemini or Copilot. This capability demonstrates the platform's potential for integrating with other advanced AI systems, further expanding the range of services and use cases that can be supported. These accomplishments mark a significant step forward in demonstrating the power of intelligent Agent orchestration in transforming government service delivery.

What we learned

We learned that Agentforce provides a powerful and flexible foundation for building multi-agent orchestration within and across Salesforce orgs. Using the out-of-the-box AI Agent Actions in Flows, we were able to easily facilitate conversations between Agents within the same org. Expanding this with the Agentforce API to support cross-org agent communication revealed the true scalability of the platform. This opened the door to a wide range of real-world government use cases—showing that with the right orchestration, Agentforce can enable intelligent, collaborative digital Agents that span departments, ministries, and jurisdictions.

What's next for CanAssist Agent

Next for the CanAssist agent is expanding beyond a single-org setup into a federated Agentforce network across multiple Salesforce orgs used by various government departments and ministries. This would be enabled via the Agentforce API and orchestrated centrally using MuleSoft as the API gateway. It will allow deeper integration with departmental systems for real-time data exchange and CRUD operations.

We also plan to:

Automate multi-step tasks, like form filling or document submission, using embedded UI/UX components.
Enable multilingual support with automatic language detection to better serve a diverse population.
Introduce predictive analytics to surface common issues and continuously optimize service delivery.
Add collaboration tools so multiple agencies and community organizations can co-own and update service content.
Enhance the employee experience by enabling Agents in different departments to easily collaborate and resolve cases across the public sector.
Ultimately, CanAssist will evolve into a limitless digital assistant that can guide, support, and complete tasks across the full spectrum of public services—efficiently and at scale.