HealthTasks Agents, Now Powered by OpenAI

For years, getting work done in SaaS meant opening the right application, navigating its screens, and learning its workflows. As AI agents become more capable, people can increasingly describe what they need and work through an AI interface instead.

Salesforce offers a useful example of this shift. Its Claudeforce partnership and its OpenAI integrations bring CRM data and workflows into conversational AI experiences. The interface may change, but useful work still depends on the trusted records and context underneath. HealthTasks faces the same opportunity in clinical education.

The interface is changing. The records still matter.

A capable AI interface can make work feel more natural, but it still needs reliable information and context to do useful work. HealthTasks is where clinical education programs manage their workflows and information. Agents can help people engage with those workflows in new ways, while HealthTasks provides the trusted context the work depends on.

This is the distinction at the heart of the next wave of SaaS: the interface people use may change, while the systems that hold an organization’s records, rules, and history remain essential.

What a harness does

Behind an AI agent is a harness: the software layer that helps the model carry out work by organizing a task, connecting it with tools and context, and managing an ongoing session. OpenAI’s managed Codex harness is one example of this approach.

We expect OpenAI and Anthropic to keep advancing this layer. We believe their harness technologies will be among the leaders for the foreseeable future, shaping how AI agents connect models, tools, and context to carry out useful work.

Over time, harnesses could become a de facto operating layer for much of the work people do online: a place to ask, coordinate, and guide work across applications. Like any useful interface, though, an agent needs access to the right systems and the right context. A conversational layer by itself cannot replace the records, workflows, and domain knowledge that make its work relevant.

Why this matters for HealthTasks Agents

HealthTasks Agents now run on OpenAI’s managed Codex harness. That gives us a modern foundation for improving how Agents work with clinical education context and respond as people guide a task.

• More efficient use of context. We are seeing improved token efficiency as Agents work through requests.
• Better follow-ups. Continue in the same thread, clarify a detail, or build on an earlier result without starting over.
• More room to steer. Guide an agent mid-run when your priorities shift or you want it to take a different direction.

For deans and program directors, this can mean a more natural way to explore program information and move from a question toward a useful next step. Faculty and staff can refine their requests as they go, instead of having to write a perfect prompt at the outset.

For IT and institutional technology teams, the harness is an important part of the architecture, while HealthTasks continues to supply the clinical education workflows and context that make the agent relevant to your program. The two layers work together: the harness helps the agent act, and the system of record gives that work its grounding.

Building for what comes next

We believe harness technologies will become a core part of how people work online. As that shift unfolds, institutions will still need systems of record that preserve trusted information, established workflows, and the context behind important decisions.

By combining HealthTasks’ clinical education platform with an evolving managed agent harness, we are working toward a future where teams can interact with their program’s information more naturally and guide AI through useful work, while keeping that work grounded in the systems they rely on.

See HealthTasks Agents

If you are exploring how AI can support clinical education workflows, we would be glad to show you how HealthTasks Agents work for your team.

Book a demo or learn more about HealthTasks.