The biggest mistake health systems can make is treating AI as another IT implementation. Over the next 5–10 years, AI is more likely to reshape how clinical work is organized than simply automate isolated tasks. The organizations that benefit most will redesign workflows, roles, governance, and incentives—not just deploy new tools.
I think there are five major shifts health systems should prepare for.
1. Move from clinician-as-documenter to clinician-as-decision supervisor
Today's clinicians spend enormous amounts of time documenting, searching for information, and coordinating care. AI is increasingly capable of handling many of these cognitive support tasks:
- Ambient documentation
- Chart summarization
- Literature synthesis
- Draft orders
- Differential diagnosis support
- Prior authorization assistance
- Patient communication
The physician's role shifts toward:
- validating AI outputs
- integrating patient context
- exercising clinical judgment
- managing uncertainty
- communicating decisions with patients
In many specialties, clinicians become "editors-in-chief" rather than primary producers of documentation.
This doesn't reduce expertise—it increases the value of uniquely human expertise.
2. Redesign work around multidisciplinary "AI-enabled care teams"
Most health systems still organize work around individual professionals.
Instead, work will increasingly be organized around teams where AI functions almost like another team member.
For example:
Current model
- Physician reviews chart
- Physician writes note
- Physician orders tests
- Nurse coordinates
- Pharmacist reviews medications
- Scheduler manages follow-up
Future model
AI continuously:
- synthesizes patient history
- identifies care gaps
- drafts documentation
- recommends evidence-based pathways
- predicts deterioration
- prepares patient education
The human team focuses on:
- complex judgment
- empathy
- shared decision-making
- ethical tradeoffs
- exceptions
- relationship building
The operating model changes from sequential work to parallel work.
3. Build AI literacy for every clinician—not just specialists
Many organizations focus on training a handful of AI champions.
Instead, AI literacy should become a core clinical competency.
Future clinicians will need to understand:
- prompt engineering for clinical workflows
- AI limitations
- calibration and confidence
- hallucination recognition
- verification strategies
- bias detection
- privacy and security
- appropriate delegation
Just as clinicians learned to interpret imaging or laboratory tests, they will need to learn how to interpret AI outputs.
4. Develop new workforce roles
The next decade will likely create entirely new clinical positions.
Examples include:
- Clinical AI implementation leads
- AI quality assurance specialists
- Clinical prompt library managers
- AI workflow designers
- Human factors experts
- Algorithm safety officers
- Clinical AI educators
- AI governance committee members
Many of these roles sit between clinical operations, informatics, quality, and digital transformation.
5. Shift from technology governance to workflow governance
Many organizations ask:
"Is this AI model accurate?"
The better question is:
"Does this workflow produce better patient outcomes safely?"
AI performance depends on:
- where it fits in the workflow
- who reviews outputs
- escalation rules
- monitoring
- accountability
Health systems will need continuous governance rather than one-time approval.
Three operating model changes
From episodic care to continuous care
Instead of interacting only during visits:
AI monitors:
- remote devices
- patient messages
- labs
- medication adherence
- risk signals
Clinicians intervene when needed instead of waiting for scheduled appointments.
From reactive staffing to predictive staffing
AI can forecast:
- patient volume
- admissions
- discharges
- staffing shortages
- ICU demand
- operating room utilization
Operations become more proactive.
From standardized pathways to personalized pathways
Clinical pathways become dynamically individualized using:
- genomics
- prior response
- comorbidities
- social determinants
- patient preferences
- real-time clinical status
Rather than one pathway for everyone, AI supports personalized recommendations while clinicians ensure they align with the patient's goals and circumstances.
The workforce challenge isn't job replacement—it's job redesign
The historical pattern with new technologies in healthcare has been that work changes more than jobs disappear.
A cardiologist in 2035 will likely still be a cardiologist.
But the workday may look very different:
| Today | 5–10 Years |
|---|
| Writing notes | Reviewing AI-generated notes |
| Searching charts | Reviewing AI summaries |
| Reading every message | Reviewing AI-triaged inboxes |
| Memorizing guidelines | Validating AI recommendations |
| Ordering routine tests | Managing exceptions and complex decisions |
| Administrative coordination | Leading multidisciplinary care |
The bottleneck shifts from information retrieval to judgment.
Organizational capabilities to build now
Health systems preparing for this future should invest in:
- Digital and AI literacy across the workforce, not just technical teams.
- Workflow redesign capabilities, including human-centered design and change management.
- Clinical AI governance, with clear policies for validation, monitoring, and accountability.
- Data infrastructure that supports interoperable, high-quality clinical data.
- Outcome measurement, tracking not only productivity but also clinician experience, patient safety, quality, equity, and trust.
A strategic perspective
The most successful health systems are unlikely to be those with the most AI tools. They will be the ones that redesign clinical work around a principle of human-AI collaboration. AI is well suited to processing information at scale, recognizing patterns, and generating drafts. Clinicians remain essential for contextual reasoning, ethical judgment, communication, and building therapeutic relationships.
Over the next decade, competitive advantage may come less from adopting AI first and more from creating operating models in which clinicians, care teams, and AI each contribute where they are strongest. Organizations that prepare their workforce through continuous learning, thoughtful governance, and workflow redesign are likely to realize greater gains in quality, efficiency, and clinician satisfaction than those that focus primarily on deploying new algorithms.
