The strategy myth: Why healthcare AI stalls at the C-suite
Overworking the need to find the perfect strategic vision is a recipe for failure for healthcare execs if there isn’t an engine to drive it.
I frequently see organizations standing paralyzed at a crossroads with respect to artificial intelligence. Organizations are waiting for a three-year roadmap, a comprehensive maturity model or a transformational strategy validated by consultants. While 75 percent of C-suite executives claim AI is a top priority, only about 1 percent have successfully scaled it.
The reason for this strategy-execution gap is simple but systemic.
We have mistaken AI for a strategy when it is actually an engine for operational wins. When we treat AI as a strategy, we fall into the paralysis-by-analysis trap. We spend months debating high-level vision while the actual plumbing of the business remains clogged. We build polished pilots that look great in a boardroom but lack the internal orchestration to survive in daily production. We are so mesmerized by the destination that we have forgotten to start the car.
Why AI efforts are stalling
The strategy first approach ironically stymies growth. It creates a missing middle or a void where the executive vision never actually connects with the operational reality. Consulting firms often deliver theoretical vs. operational success. You see a demo of what’s possible then leave without the functional engine required to move that pilot into the real world.
If your organization is stuck, it is likely because you are trying to do the same things more efficiently rather than reimagining the process itself. Strategy focuses on optimization while treating AI as an engine enabling transformation.
Focused AI efforts
To break paralysis by analysis, we must shift the focus to operational wins. Start picking projects where ROI can be proven immediately. Don’t ask AI to perform surgery; instead, focus AI experimentation efforts on a set of interconnected functions or tasks within a single domain like claim resolution. One-off disconnected use cases don’t improve workflow and processes.
A BCG study updated in 2025 found that firms using the deep and narrow approach have twice the ROI of firms using shallow and broad deployments. Specifically, the study highlights a core BCG principle often referred to as the "10-20-70 Rule" for AI success. The rule suggests that 10 percent of the firms’ effort is the algorithm (the AI tool itself); 20 percent of the effort is the technology/data (the engine/infrastructure); and 70 percent of the effort is people and process change (change management).
Learning from history
When Lou Gerstner arrived at IBM in 1993 from RJR Nabisco, the company was in serious trouble, including being seen as outdated in the PC era. His famous stance, “The last thing IBM needs right now is vision,” wasn’t anti-strategy. It meant that IBM didn’t need abstract future-state thinking. IBM needed execution, focus and survival discipline. IBM needed operational wins. Gerstner engineered one of the greatest corporate turnarounds in history.
For healthcare organizations applying this lesson, that means starting by operationalizing AI project by project and start counting wins and improved ROI.
A prime example is the national provider identifier (NPI) enrollment process. Historically, this is a tedious, manual task. By treating AI as a tool at the orchestration layer, we collapsed a 10-day process into just two hours. The result was immediate utilization of hundreds of NPIs to obtain coverage and benefit information rather than waiting for a prolonged current state manual enrollment process.
This wasn't achieved through a grand strategic shift, but through agentic AI orchestration.
The orchestration layer acts as the smart engine or operation system that manages the workflow in real-time while communicating to sub-agents.
Subagent specialization was achieved by deploying specialized agents or automated workers that handle specific tasks like geographic tracking of NPIs for enrollment purposes, payer communication and database tracking including NPIs requiring a wet signature like NY Medicaid.
Goal-oriented reasoning agents are unlike traditional automation (RPA), which follows a rigid script; by contrast, goal-oriented reasoning agents can reason. If a payer database is down, the agent doesn't just error out; it flags the issue and retries. It solves the problem functionally and not strategically.
Flattening the J-curve
Standard industry research shows a productivity J-curve is a dip in performance when a new strategy is plugged in without redesigning the work. By focusing on wins project by project, we effectively flatten that curve.
We aren't waiting for a three-year transformation. We are proving that AI is a capability that delivers results and confidence today. The win isn't in the roadmap – it’s in the execution.
If you are a leader in health data management, my advice is to stop asking for a more detailed roadmap and start building a better engine. Pick one high friction process, deploy an agentic layer, and turn days of work into hours.
I’ll close with a quote from my favorite author, Dr. Seuss. “Don’t call it AI; that’s a bit of a fib. It’s like saying a fork is a dinnertime bib. Keep your eyes on the goal. Keep your path in your sight. And then use AI to go left or go right.”
Ken Poray is CEO of Integrex Health and Chair of the AI Community of Practice at the American College of Health Data Management. He has 20 years of experience working with payers, status, EDI transactions and most recently with AI workflows.