Artificial intelligence has become a cornerstone of innovation in healthcare, promising improved efficiency, better outcomes, and reduced clinician burnout. Yet despite massive investments, most organizations capture only 20 to 25 percent of AI’s potential value. The reason? It’s not the technology—it’s the last mile of adoption.
In Episode 19 of the Impactful AI podcast, host AJ sat down with Claude Opus 4.1 to explore why so many AI initiatives stall after deployment, and what healthcare leaders can do to close the gap between potential and performance.
The Real Challenge Isn't Building AI - It's Using It
As AJ pointed out, the hardest part of AI isn’t building the model or deploying it—it’s what happens after the tool goes live. That final stretch, where AI is integrated into real workflows and used consistently by clinicians and staff, is where most projects falter.
Claude emphasized that last mile adoption is often misunderstood. Organizations focus on technical readiness but overlook the human and operational factors that determine whether AI tools actually deliver value. It’s not just about fitting AI into existing workflows—it’s about reimagining those workflows and aligning incentives across the organization.
Understanding the "WIIFM" Factor
One of the most powerful insights from the episode was the importance of clarifying the “WIIFM”—what’s in it for me. AI scribes, for example, succeeded because clinicians immediately saw the benefit: “I get my evenings back.” That’s far more compelling than abstract metrics like “12 percent documentation time saved.”
Claude explained that different stakeholders have different motivators. A night-shift nurse might value predictive analytics because it boosts confidence during coverage. A manager might appreciate it for justifying resource requests. These motivators must be surfaced early and clearly communicated to ensure adoption.
Activated Shelfware: When AI Goes Unused
Without thoughtful last mile planning, organizations risk ending up with “activated shelfware”—tools that are technically live but rarely used. Claude shared examples of AI systems that generate discharge instructions or predictive alerts, but fail to deliver value because no one consistently integrates them into care.
Even enterprise platforms like EHRs and ERPs now come bundled with AI features. But flipping the switch isn’t enough. Without training, workflow redesign, and user engagement, these tools remain underutilized, creating an illusion of progress without real impact.
The Cost of Neglecting Adoption
The statistics are sobering. Across industries, organizations capture only 10 to 30 percent of AI’s potential value. Nearly half of companies abandoned most of their AI initiatives in the past year. But there’s a silver lining: the 16 percent that invest in last mile adoption—budgeting for support, shadowing users, and aligning incentives—see higher revenue growth.
Practical Advice for Healthcare Leaders
So what can healthcare executives do to close the last mile gap? Claude offered three actionable strategies:
- Harvest Value, Don’t Just Buy Tools
Keep exploring new AI capabilities, but also focus on extracting value from existing investments. Many organizations already have AI features embedded in their platforms—make sure they’re being used effectively.
- Make Ownership Explicit
Adoption must be someone’s job. Whether it’s a new role like an AI adoption lead or assigning responsibility to an existing operational leader, clear accountability is essential. If it’s everyone’s job, it’s no one’s job.
- Budget for Adoption
Spending a million dollars on AI without allocating funds for adoption efforts—like training, workflow redesign, and measurement—is a recipe for underperformance. Even dedicating 20 percent of the budget to adoption can significantly boost ROI.
Andrew added that organizations should plan for a three to six-month adoption arc. It starts with handholding and shadowing, then moves into habit-building. Real change takes time, and AI is no exception.
The Bottom Line: Execution is the Differentiator
The AI divide today isn’t about access to technology—it’s about execution. Most organizations have similar tools. The winners will be those that do the unglamorous work of last mile adoption: engaging users, redesigning workflows, and aligning incentives.
As Claude put it, “AI without the last mile is expensive decoration. With it, it can be truly impactful.”
