At TDX 2026, the Agentforce Hackathon brought together builders from across the Salesforce ecosystem to create AI-powered solutions that tackle real-world challenges.
And the ideas were anything but theoretical. This year’s finalists explored how Agentforce can:
Along the way, they also uncovered something bigger: Building with AI agents requires a different mindset than traditional automation.
We asked the three finalists from the TDX Hackathon Showdown to share what they learned while building with Agentforce, what surprised them most, and what advice they’d give others experimenting with AI-led experiences in Salesforce.
Gigi Chan is a Technical Architect and Slack Community Leader focused on building AI-driven experiences that improve operational efficiency and accessibility.
Her solution, CareRoute, uses Agentforce to help care coordinators process referrals, match patients with the right clinicians, and identify patients at risk of being lost to follow-up — all while supporting multilingual communication workflows for more accessible healthcare experiences.
Eleanor Spolyar and Taryn Kelley partnered to create City Pulse Agent, an AI-powered solution designed to help cities move from reactive service management to proactive infrastructure planning.
By combining Agentforce, Slack, Tableau, and Agentforce Service, City Pulse analyzes historical infrastructure data, identifies trends, and helps city teams coordinate preventative maintenance more effectively through collaborative, real-time workflows.
Arundathi Neelam is a Senior Salesforce Analyst with 5 years in the Salesforce ecosystem and experience across Sales Cloud, Health Cloud, and Marketing Cloud. Previously at IBM and Accenture, she now focuses on exploring autonomous experiences with Agentforce.
Her solution, KnowledgeForce, is a self-healing Agentforce knowledge management system for global trade compliance. It’s a knowledge base designed to build and maintain itself through contextual understanding, reasoning, and conversational AI experiences.
Gigi Chan, CareRoute: The user interface for building Agentforce provides an excellent development experience and makes it easy to quickly prototype and refine AI-powered workflows. The CareRoute agent was developed using Agent Script, which enabled more deterministic agent instructions and improved the consistency and accuracy of outputs. During development, the built-in troubleshooting agent within Agentforce Builder was especially valuable for diagnosing and refining Agent Script behavior. Event logs also made it easy to review past conversations, trace agent actions, and efficiently debug issues. Overall, the platform feels intuitive and approachable, making it accessible for both technical and non-technical users.
Eleanor Spolyar and Taryn Kelley, City Pulse Agent: Building with Agentforce feels way less rigid than traditional automation. Instead of just building “if this, then that” workflows, you’re designing something that can actually understand context, respond conversationally, and coordinate actions across multiple systems in real time. It feels more like designing a digital operations team member than a standard app.
Arundathi Neelam, KnowledgeForce: With traditional automation, you’re scripting every path the system can take. With Agentforce, you’re giving an agent goals and trusting it to reason through how to get there. It’s a real mindset shift — less ‘what happens when X’ and more ‘what should the agent know and decide.’ Once that clicks, you start designing for intent instead of clicks.
Gigi Chan, CareRoute: The major challenges I addressed were intelligently matching patients to the right clinicians, improving multilingual patient communication, and identifying patients who may be at risk of being lost to follow-up. Traditional automation alone could not fully solve these problems because healthcare referrals require understanding nuance, context, and the many ways the same condition can be described, rather than relying on rigid rules or keyword matching. With Agentforce, I was able to support more accurate clinician matching, generate follow-up emails in a patient’s primary language, and analyze clinical notes for indicators that a patient may need additional outreach, helping care coordinators intervene earlier.
Eleanor Spolyar and Taryn Kelley, City Pulse Agent: Traditional automation is great for handling routine tasks, but it doesn’t always help with identifying larger patterns or predicting potential future issues. City Pulse solves the challenge of helping cities move from a reactive approach to a more proactive one. City Pulse analyzes trends from service requests and infrastructure data to anticipate problems before they occur. Traditional automation can streamline workflows, but it doesn’t provide the insights needed to prevent issues altogether.
Arundathi Neelam, KnowledgeForce: Knowledge bases go stale the moment they’re published — and no flow can tell you what is missing or outdated. KnowledgeForce flips that. Every unanswered question becomes a signal that the knowledge base has a gap, and the system drafts a new article to fill it. It’s something declarative automation alone couldn’t do, because the trigger isn’t a record change — it’s a reasoning failure.
Gigi Chan, CareRoute: Slack and Data 360 had the biggest impact on my solution because they addressed two key needs: improving care coordination and bringing unstructured healthcare data into Salesforce in a usable way. Slack helped streamline coordinator workflows by reducing the need to switch between Salesforce, electronic health record (EHR) systems, and other tools, while Slackbot made it easier to pull information from referral PDFs and Salesforce records, verify extracted data, and validate whether Agentforce recommended the right clinician match. Data 360 was also important because Document AI transformed referral PDFs into structured data and ingested clinical notes from external EHR systems, which prompt templates then converted into structured information Agentforce could analyze to identify patients who may be at risk of being lost to follow-up.
Eleanor Spolyar and Taryn Kelley, City Pulse Agent: Tableau had a major impact on the City Pulse solution because it helped us visualize all of the data in a digestible way. It turned complex data into interactive visuals showing trends and insights that may have been missed. At the same time, Tableau elevated the approach from reactive service management to predictive operations. By showing real-time trends and patterns, this helps city leaders spot potential issues early and address them before they become bigger problems.
Arundathi Neelam, KnowledgeForce: The combo of Agentforce and Flow. Agentforce handled the reasoning and conversation, but Flow did the heavy lifting behind the scenes — routing drafts for review, updating records, and triggering Slack notifications. It reminded me that AI agents are most powerful when paired with solid declarative plumbing, not used in isolation.
Gigi Chan, CareRoute: One thing that surprised me most was Agentforce’s ability to seamlessly support multilingual communication. When care coordinators draft follow-up emails for patients, I wanted the message to be generated in the patient’s primary language to improve accessibility and reduce the risk of miscommunication. At the same time, I needed the draft to also be displayed in English so the care coordinator could review and validate the content before sending it. Agentforce was able to handle both simultaneously, which made the workflow much more efficient while still maintaining oversight and quality of care.
Eleanor Spolyar and Taryn Kelley, City Pulse Agent: We were honestly surprised by how natural it was to interact with the agent once it was up and running. The experience immediately felt more intuitive than having someone fill out a traditional form or navigate a portal.
We were also surprised by how fast we could stand up an end-to-end experience using different parts of the Salesforce ecosystem. Connecting Agentforce, Slack, Tableau, and Agentforce Service together happened much faster than we expected, which let us focus more on the experience itself versus just wiring technology together.
Arundathi Neelam, KnowledgeForce: How much of the work is prompt design, not platform configuration. The moments that made or broke the experience were almost always in the prompt template. Tiny wording changes shifted the agent’s behavior more than any flow tweak.
Gigi Chan, CareRoute: I would recommend starting by exploring the actions and sub-agents that are already available in Agentforce. New capabilities are being added constantly, and many common use cases are already supported out of the box, so you do not always need to build everything from scratch. Using prebuilt components makes it easier to prototype quickly, validate ideas, and focus on refining the workflow, user experience, and business value.
Eleanor Spolyar and Taryn Kelley, City Pulse Agent: Start with a real business problem or user interaction instead of starting with the AI itself. The best Agentforce experiences come from identifying inefficiencies in an existing process and figuring out how the agent can make that experience faster, easier, or more natural.
Also, don’t overengineer it upfront. Build something quickly, test real conversations, and iterate from there. We learned more from testing live interactions than we did trying to perfect everything on paper first.
Arundathi Neelam, KnowledgeForce: Start smaller than you think. Pick one task an agent could realistically own end-to-end, build it, and watch where it breaks. You learn way more from one fully working agent than from a sprawling one that half-works. And test with messy, real questions — not the clean ones you’d put in a demo.
One thing was clear across every finalist solution: Builders across the Salesforce ecosystem are already shaping the future of AI-powered work with Agentforce.
Whether they were improving healthcare coordination, helping cities anticipate infrastructure issues, or creating more contextual knowledge experiences, these teams demonstrated how AI agents can go beyond traditional automation to deliver more intelligent, adaptive, and human-centered experiences.
And perhaps most importantly, they showed that the future of building with AI isn’t just about faster workflows. It’s about designing systems that can reason, collaborate, and support people in entirely new ways.
Ready to start building? Explore Agentforce and see what you can create.
Salesforce Headless 360 is starting to show up in conversations with your developers and leadership. If you’ve been wondering what it actually means for your day-to-day life, or whether it changes anything you do, you’re in the right place. The short answer: It makes the work you already do more powerful and more reusable. Your […]
We’re standing at the edge of the biggest shift our industry has ever seen. The Agentic Enterprise is transforming how businesses operate and how we work. Salesforce Admins aren’t observing this shift from the sidelines. You’re building it, piece by piece, every day. This is the natural evolution of a community that spans 160+ user […]
TDX 2026 returns to San Francisco and Salesforce+ on April 15–16. This year we’re introducing brand-new experiences designed to help you go deeper on the Agentforce 360 Platform and connect directly with the experts building it. TDX is where admins, developers, and architects come together to sharpen technical skills and have real conversations about what’s […]
