David Lareau, president and CEO of Medicomp Systems
Photo: David Lareau
The next wave of healthcare's adoption of artificial intelligence will focus more on collaboration, with the model context protocol helping large language models work in tandem with other algorithmic tools to bring the industry closer to true voice-driven interaction in healthcare, said David Lareau, president and CEO of Medicomp Systems, a clinical IT company.
The model context protocol is an emerging industry standard that defines how AI systems, large language models and agent-based applications connect with trusted knowledge sources.
Cleaner documentation and smaller AI models
This is just one of Lareau's predictions for health IT in the year ahead. He also has his eyes firmly set on payers and providers embracing tools that deliver clean and thorough documentation that drives accurate reimbursement and better patient outcomes, and on healthcare organizations turning to smaller, domain-specific AI models.
"When I look at how organizations have adopted AI over the past few years, it is clear that many enterprises remain grounded in large, monolithic systems," he observed. "The model context protocol changes the equation by providing systems with a standardized way to communicate with AI agents that perform highly specific tasks.
"It lets developers focus on building purpose-driven tools without requiring deep integration into the host electronic health record," he continued. "That creates a more open pathway for collaboration across the industry."
Healthcare is seeing early momentum in areas such as voice-driven interaction, he added.
More flexibility for innovation
"Many companies are developing ambient listening and voice-command technologies that enhance documentation and improve the user experience," he said. "These offerings can connect to enterprise platforms through defined APIs supported by the MCP. That reduces the need for custom code and supports a more flexible ecosystem for innovation.
"No single vendor can meet every organizational need, especially as AI capabilities diversify," he continued. "Health systems want the freedom to adopt best-of-breed platforms that address focused use cases while still fitting into their core workflows. The MCP provides the structure to make this possible by connecting targeted AI capabilities to enterprise systems in a predictable way. This brings us closer to more natural and preferred voice-driven support for clinicians at the point of care."
On another front, as Medicare audits intensify, Lareau predicts payers and providers will seek to strengthen the quality and completeness of their data and embrace tools that deliver clean and thorough documentation that drives accurate reimbursement and better patient outcomes.
"We have made real progress in capturing information through tools such as ambient listening, but that progress increases the responsibility to validate what is documented," he said. "As Medicare Advantage enrollment grows, audits are focusing on whether documented diagnoses truly reflect conditions that are present and being managed.
"That scrutiny will continue, and organizations will need to verify documentation with greater precision," he continued. "If they do not, they face significant financial and compliance risk."
Appropriate clinical evidence
Lareau expects to see tools become available that can review encounters and charts to confirm that diagnoses are supported by the appropriate clinical evidence. These tools can help providers validate documentation in real time and help enterprises conduct deeper reviews across broader populations, he said.
"Payers will also use these capabilities to confirm accuracy before submitting data to Medicare," he explained. "This creates a shared need across the healthcare ecosystem for reliable validation to support sound clinical decision making and protect organizations from significant downstream risk.
"The industry's push for stronger documentation accuracy and clearer clinical support for diagnoses aligns with the evolution of the CMS Hierarchical Condition Category model, which now requires greater specificity in coding to improve risk-adjustment accuracy," he added. "The latest version of the model demands greater clinical detail and raises the bar for documentation quality, which reinforces the need for strong validation processes."
As the transition to the new model completes in 2026, organizations will depend on technologies that surface documentation gaps and confirm clinical accuracy, he said.
"These capabilities will support appropriate reimbursement and better patient outcomes by helping organizations confirm that every documented condition reflects the clinical reality of the encounter and that care plans address the patient's true needs," he added.
Smaller AI models are more cost-effective
Finally, Lareau says as organizations scale large language models across enterprise systems, leaders will increasingly turn to smaller, domain-specific AI models that are more cost-effective, and which can operate securely within their own environments.
"When health systems explore how to use AI at scale, then cost, security and operational reliability become central considerations," he explained. "Large language models require significant computing resources and introduce token consumption costs that escalate quickly when thousands of clinicians rely on them during routine care.
"Smaller, domain-specific models present a more practical option because they can operate on standard CPUs, rather than GPUs, and can be deployed within a health system's own protected environment," he concluded. "That helps organizations manage expenses while maintaining control over sensitive clinical data."
Follow Bill's health IT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.
WATCH NOW: In 2026, AI will power change in RTLS
