AI continues to hold promise for transforming healthcare operations, but realizing that potential requires more than plug-and-play solutions. In a featured session at Becker’s Fall 2025 Payer Issues Roundtable, leaders from Raport and Precision Medicine Diagnostic Partners explored what it takes to build reliable, repeatable AI workflows — and how deeper collaboration between engineers and clinicians can unlock scale.
Here are five key takeaways from the discussion:
1. Understanding what AI agents are — and are not — is essential.
AI agents are often misunderstood in the healthcare space. According to Nathan Feldt, chief technology officer and co-founder of Raport, true agents interpret context, plan, use tools and act autonomously toward a defined goal. Many current tools on the market, however, are rigid workflows masked as agents.
“What you’re expecting is that the agent can come up with its own solution to the problem,” Mr. Feldt said. “In contrast, a workflow is more predefined — you know exactly what the criteria are and the steps to meet them. An agent generates its own solution, whereas with a workflow, you provide the steps. Most of the products out here today are workflows, not agents.”
This distinction matters, he noted, because workflows, especially when built in low-code or no-code environments, lack the rigor needed for complex clinical use cases, where reliability and repeatability are paramount.
2. Fragmentation is slowing AI’s ability to scale.
Mr. Feldt explained that specialized AI vendors often build narrowly scoped tools that operate in silos, which creates technical and operational inefficiencies. “Managing these systems adds complexity, and honestly takes away a little bit of the benefits that you’re getting from these systems when you’re having to switch off to all these different integrations,” he said.
Mauricio Garcia Jacques, MD, CEO and co-founder at Raport, added that context retrieval — a key function that powers clinical AI workflows — often gets redundantly built across vendors, wasting resources and leading to incompatible data models. “Ultimately, what we hope to avoid is rebuilding infrastructure only to achieve similar capabilities between these systems,” he said.
3. A unified infrastructure can enable more intelligent, scalable automation.
Raport’s platform aims to solve these challenges by creating a common ecosystem of tools, subagents and workflows that share context and outputs. One example: the output of an intake summarization workflow can feed into downstream processes like prior authorization — simplifying documentation and reducing duplication. This approach, Mr. Feldt said, enables the company to build affordable custom solutions that scale.
“We believe AI has the capability to transform healthcare,” Mr. Feldt said. “But the current landscape of these systems makes it incredibly difficult to scale. Our goal at Raport is to be able to build a system that allows for performance and scalable AI systems that also support interoperability and avoid this utility fragmentation that we’ve been talking about.”
4. Cancer care’s complexity demands better data and workflow alignment.
Precision Medicine Diagnostic Partners grounded the conversation in the realities of modern oncology: rising cancer incidence, expanding biomarker requirements and care journeys that involve numerous specialists, tests and reports. Co-founders Judy Largen, BSN, RN, and Dot Guccione, MSN, RN, emphasized that inconsistent data capture and coordination often stretch diagnosis and treatment timelines from an ideal two weeks to nearly two months.
Their team has built analytics algorithms, specialized data models and visualization tools to surface testing gaps, bottlenecks and guideline deviations. By translating complex, real-time clinical data into clearer insights, they aim to help health systems identify where delays occur and support faster, more precise cancer care.
5. AI-clinician partnerships are essential to building clinically meaningful automation.
Precision Medicine Diagnostic Partners and Raport are working together to pair deep oncology expertise with scalable AI infrastructure; while Precision Medicine Diagnostic Partners contributes domain knowledge and clinical data methodologies, Raport provides a workflow engine capable of operationalizing those insights across care teams and systems. The goal is to build interoperable, extensible workflows that reflect real-world cancer care and can expand to other diseases.
“Our collective intent is to collaborate on our methodologies, technologies and solutions because we really do want to identify those patterns in the data that help us point to specific gaps in cancer care,” Ms. Largen said. “We believe this partnership will produce those streamlined workflows and strong clinical and patient support tools, and expand applicability to additional diseases, which is really exciting because this can be replicated into rare disease, chronic disease and more.”
As a word of caution to healthcare leaders, she added that human insight remains essential: “AI is so valuable, but human insight is critical. You can’t forget that.”
