David Lareau, CEO and president of Medicomp Systems
Photo: David Lareau
At the 2026 HIMSS Global Health Conference & Exposition last week in Las Vegas, David Lareau, CEO and president of Medicomp Systems, a vendor of evidence-based, AI-powered clinical intelligence systems, said a key development facing attendees is the rapid enterprise deployment of generative AI and ambient documentation tools directly into clinical workflows. Especially, he added, where AI-generated content is entered into the patient medical record.
"Health systems are moving beyond experimentation and embedding large language models into documentation, chart summarization, coding support and clinical decision workflows," he noted. "The issue is that these models are probabilistic and predictive in nature, which introduces variability and potential inaccuracy into clinical documentation.
"When AI-generated outputs are accepted without validation, unsupported diagnoses, incomplete clinical context and subtle inaccuracies can become part of the permanent record," he continued.
Safety and integrity
This trend carries direct implications for patient safety, reimbursement integrity, quality reporting and regulatory compliance, he said.
"An AI model may infer a condition based on incomplete contextual cues, summarize findings in ways that omit clinically relevant negatives, or generate documentation that appears coherent but lacks evidentiary support in the chart," Lareau explained. "Because LLMs can produce different outputs from identical inputs, clinicians and organizations may have limited visibility into when information has drifted from the clinical truth.
"As adoption accelerates, the downstream impact on coding accuracy, risk adjustment and care continuity becomes more pronounced," he continued. "The broader industry conversation is, therefore, shifting toward responsible AI, governance and clinical validation frameworks. Healthcare organizations are recognizing that speed and usability must be balanced with determinism, transparency and traceability."
The challenge for 2026 is how to harness AI-enabled efficiency while preserving clinical integrity – that balance will define the next phase of AI maturity in healthcare, he added.
Put up the guardrails
The term "clinical guardrails" is ubiquitous at this point, but it does carry real meaning for CIOs and health IT leaders evaluating AI adoption, Lareau said.
"In a clinical documentation context, these guardrails need to evaluate and validate AI-generated outputs before they are committed to the medical record or used to drive billing and quality decisions," he explained. "Implementing this guardrail requires integrating deterministic, evidence-based intelligence alongside generative models to assess whether suggested diagnoses, problem list entries or clinical summaries are supported by documented findings.
"Validation must occur at the point of creation, with transparency into why a condition is or is not clinically substantiated," he continued. "Without this layer, organizations assume greater clinical, financial and regulatory risk."
Leaders should also evaluate their data foundation and terminology infrastructure – AI systems operate on the data they receive, so fragmented, poorly mapped or inconsistently coded inputs amplify the risk of inaccurate outputs, he added.
Normalizing inputs
"A comprehensive clinical data foundation and knowledge graph can normalize inputs across domains, apply diagnostic relevancy logic, and ensure that outputs align with established clinical evidence," he stated. "This structured layer becomes essential for supporting risk adjustment, quality measures, Hierarchical Condition Category coding and interoperability requirements.
"In parallel, CIOs must address privacy architecture and cost control," he continued. "Minimizing exposure of protected health information to external models and leveraging architectures that transmit structured queries rather than full narrative data can significantly reduce both compliance risk and token-consumption costs."
As models grow larger and usage expands, operational expense management becomes strategic, potentially requiring a switch to smaller, more targeted language models, he added.
"Organizations that combine validated intelligence, secure architecture and cost discipline will be best positioned to scale AI responsibly," he concluded.
Medicomp Systems is in booth 5435 in the exhibit hall at HIMSS26.
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: Grabbing the Chief AI Officer brass ring – and working with top brass
