Be cautious but move quickly: 6 actions to successfully deploy AI in healthcare

It’s only been a year since Senior Intelligence Director Andrew Rebhan last took the stage at Sg2 Executive Summit to discuss the applications of artificial intelligence in healthcare — but in AI years, that’s a lifetime.

Just 12 months ago, many organizations were still grappling with basic AI definitions. “Now we’re talking about systems of agentic AI that are incorporating multiple different personas under one platform,” Rebhan told the audience during a panel discussion titled Advancing Health Care Performance: Real-World Applications of AI at this year’s Summit. “That is a massive scale of learning we’ve accomplished in the span of a year.”

Even in healthcare, the pace of AI development has transitioned from theoretical to tactical at breakneck speed. From ambient scribing technologies that reduce physician burden to AI-driven revenue cycle automation and streamlined supply chain management, health systems are finally seeing tangible use cases move from pilot to scale.

It’s no longer about exploring what AI could be. It’s about executing what AI can do.

This shift is reflected in both tone and tactics: Ambient solutions like those used for clinical documentation are widely viewed as the “golden application” of AI in 2025, but they’re now joined by a host of back-office innovations aimed at unlocking efficiency and reducing administrative waste. It’s no longer just about saving time — it’s about transforming the workforce, redesigning operations and extending capacity in ways that align directly with healthcare’s biggest challenges.

Of course, approaching AI implementation with the right strategy is central to unlocking the value propositions outlined above. That’s why during the panel Rebhan and Vizient Senior Vice President of New Ventures Robert Lord described key actions to consider when integrating AI into organizational workflows, including …

1. Align AI with strategy and outcomes

Too often, AI projects begin as siloed tech experiments with unclear value. Lord and Rebhan both emphasized the need to treat AI not as a novelty but as an enabler of strategic priorities.

Key actions:

• Establish enterprise-level AI strategy: Build a top-down vision that identifies clear goals such as improving access, reducing clinician burnout or increasing revenue.
• Use a maturity model: Lord discussed a multi-dimensional maturity framework that includes:
  • People and culture: Are staff AI-aware or AI-skilled?
  • Data and tech infrastructure: Are your systems interoperable and secure?
  • Measurement and ROI: Do you track both financial and experiential outcomes?
  • Strategic alignment: Are all projects clearly tied to organization-wide goals?
• Track value beyond dollars: Rebhan outlined a five-pillar framework:
  • Financial impact, such as increased collections and fewer denials
  • Operational efficiency like automation of back-office workflows
  • Clinical outcomes, including predictive modeling for improved care risk management
  • Workforce experience, such as less time charting and lower burnout
  • Access improvement, including reduced wait times and streamlined communication.

2. Redesign governance to build trust, not bottlenecks

“Governance is not about saying ‘no’ — it’s about creating systems that earn trust.” — Robert Lord

Traditional IT governance structures don’t fit the agility demands of AI. Lord called for modernizing governance to balance innovation with responsible oversight.

Key actions:

• Move beyond checklists: Shift from reactive approvals to strategic, dynamic governance that enables experimentation.
• Create multidisciplinary governance teams: Involve IT, compliance, business leaders, clinicians and end-users in decisions.
• Align on shared objectives: Governance must reinforce enterprise goals and ensure that each AI use case supports organizational strategy.
• Use AI to govern AI: Tools can monitor compliance, bias and usage patterns in real time, creating self-auditing systems that reduce risk and increase transparency.

3. Experiment intentionally: Start low-risk and scale smart

Rather than trying to tackle high-stakes clinical AI from the start, take a stepwise, risk-calibrated approach: Think “moonshot readiness” through low-orbit tests.

Key actions:

• Launch in operational domains first: Revenue cycle management, scheduling bots or supply chain offer low clinical risk but high value.
• Use these early wins to build cultural confidence: Demonstrating fast, measurable success helps shift AI from skepticism to excitement.
• Be pragmatic, not trendy: Rebhan advises not to chase every new model or vendor. Focus your resources where AI has clear impact and data readiness.
• Institutionalize prioritization: Only 60% of health systems in a recent Vizient survey reported having a formal AI prioritization process — this is a must-have to avoid scattershot pilots.

4. Move from pilots to enterprise execution

“We’re seeing many organizations with tons of pilots and no framework for prioritizing.” — Andrew Rebhan

Both experts warned against pilot fatigue and the temptation to spread resources too thin. Momentum and value come from scaling strategically.

Key actions:

• Consolidate learnings from existing pilots: Systematically evaluate outcomes and identify what’s ready to scale.
• Fund AI through strategy, not IT scraps: Rebhan noted that most organizations still draw AI funds from IT budgets, but real transformation requires dedicated, strategic investment.
• Choose your vendor approach wisely:
  • A platform-first approach offers speed and integration.
  • A best-of-breed approach can unlock niche innovation but may increase complexity.

5. Rewire the workforce for AI readiness

“We can rewire our workforce to be more engaged and work at the top of their license.” — Robert Lord

AI isn’t just a technology shift: It’s a workforce transformation. Lord emphasized that empowering staff is essential to realizing AI’s full potential.

Key actions:

• Conduct an AI-readiness audit: Assess your team’s current capabilities, openness and needs around AI integration.
• Invest in upskilling: Develop internal training to raise AI fluency — especially for clinical and operational leaders.
• Involve clinicians early: AI projects succeed when clinicians feel heard and involved, not when solutions are imposed from the top.
• Frame AI as augmentation, not replacement: Reinforce how AI helps staff focus on higher-value tasks and improve care.

6. Treat urgency as a competitive imperative

“If you go too slow, you ironically put yourself on a fast track to obsolescence. By that point, the only thing slower than your AI rollout is going to be your recovery in the competitive market.” — Andrew Rebhan

Healthcare has a reputation for lagging in tech adoption. But in this AI cycle, everyone is still early — and that’s an opportunity.

Key actions:

• Recognize your advantage: Even large tech companies are figuring out AI implementation. Health systems that act with urgency can leap ahead.
• Balance speed with stewardship: You don’t need to rush recklessly, but delay by default is no longer defensible.
• Start somewhere — today: Even one well-scoped, well-governed use case (like AI for denial management or physician note automation) can catalyze cultural momentum. “Demonstrating fast, measurable success helps shift AI from skepticism to excitement,” Lord said.

Learn more about our Data and Digital expertise and solutions.