As healthcare organizations race to implement artificial intelligence (AI), it’s critical to understand that AI encompasses a wide spectrum of technologies, and each deployment needs an individual strategy.
Rob Havasy
Senior Director, Informatics Strategy, HIMSS
Successfully deploying AI in today’s complex healthcare landscape is predicated on three pillars: understanding the AI solution, determining your organization’s data maturity and readiness for integration, and establishing stakeholder trust through transparent, ethical leadership.
Defining your AI solution
The term “AI” often obscures more than it clarifies. At one end of the continuum, narrowly focused systems tackle individual tasks such as radiology image triage and automated prior-authorization workflows. On the other hand, expansive platforms offer cross-functional reasoning in clinical, operational, and administrative domains.
Key dimensions include:
- Clinical vs. Administrative: A clinician-facing diagnostic support tool requires performance benchmarks and validation protocols different from those of a revenue-cycle optimization engine.
- Internal vs. External Training: In-house models, trained on your patient population, can deliver heightened specificity and sensitivity; by contrast, prepackaged solutions enable rapid deployment but may underperform in niche contexts.
- Static vs. Self-Learning: Fixed-state models maintain consistent behavior against a defined dataset, whereas adaptive (agentic) systems evolve, amplifying the need for ongoing validation and governance.
Tailoring integration, validation, and governance processes to these distinctions is non‐negotiable; there is no universal blueprint.
Aligning on data, interoperability, and governance
“You cannot have AI governance without data governance,” observed a HIMSS member during a recent panel discussion, a sentiment that resonates across the industry.
Organizations vary widely in interoperability, data cleanliness, and governance maturity. Each dimension directly influences implementation complexity, which is why our work at HIMSS, particularly our suite of maturity models, revolves around helping health systems benchmark their systems and bridge the gap between technology and healthcare.
AI solutions can access live data streams with minimal friction in highly integrated environments powered by modern EHR platforms and robust FHIR APIs. Yet many health systems still rely on legacy databases and siloed applications, necessitating an interim ETL (Extract, Transform, Load) process to normalize and stage data for analytic pipelines.
While this intermediary step accelerates short‐term AI deployment, sustained efficiency and reliability require ongoing integration and data management investment.
Cultivating stakeholder trust
Technical prowess alone is insufficient; human buy‐in is the key to successful AI deployment.
Clinicians and administrative staff must trust that AI tools are accurate, interpretable, and genuinely augment their expertise rather than threaten their roles. Establishing clear governance frameworks detailing model training, validation, performance monitoring, and bias mitigation is essential.
Leadership must communicate candidly about workforce impacts, champion upskilling programs, and reinforce that AI is an intelligence amplifier, preserving professional judgment rather than supplanting it.
AI has the potential to impact all levels of an organization, and all healthcare professionals should be in touch with their peers to share best practices and lessons learned. At the annual HIMSS AI in Healthcare Forum, which is coming up in July in New York, we’ll come together for a deep exploration of critical implementation considerations, and guests will leave better equipped to deploy AI responsibly, ethically, and effectively, advancing patient outcomes and organizational performance.
Care doesn’t exist in a vacuum, and we must look to one another for help in successfully integrating AI across clinical, operational, and administrative domains.
