Healthcare’s AI Problem Is Not Pilots. It’s Workflow Design.

Healthcare does not have an AI experimentation problem. It has a workflow design problem.

That distinction matters because most health systems already believe AI matters. They are testing note-writing tools, contact center assistants, prior-authorization automation, and staffing support. But too much of that activity still sits in the category of pilot work. It gets layered onto legacy processes instead of forcing a redesign of the work itself. Deloitte’s 2026 State of AI in the Enterprise says the direction is consistent: organizations should take an AI-native approach and redesign work holistically rather than simply adding AI to existing systems.

That is the real strategic issue in healthcare right now. The sector does not need more fragmented experimentation. It needs executive teams willing to redesign frontline operations around agentic execution.

Pilot activity is not the same thing as operating change

Healthcare has no shortage of AI motion. Hospitals, health systems, and care networks are testing AI for documentation, triage support, staffing coordination, and patient communications. On the surface, that sounds like progress. In reality, many of these efforts still behave like overlays. The workflow stays the same. The approvals stay the same. The routing logic stays the same. The human bottlenecks stay the same. AI just gets inserted into one narrow step.

That approach limits value. It creates localized efficiency gains without changing the operating system underneath. The result is usually a familiar pattern: executives can point to pilot activity, but frontline teams still feel the same friction in handoffs, staffing strain, escalation paths, and queue management.

McKinsey’s latest work on frontline nursing makes the stakes clearer. The opportunity is not just better documentation or marginal time savings. It is the ability to rethink how work gets distributed, supported, and escalated so clinicians spend more time on higher-value care and less time inside repetitive administrative loops.

The bottleneck is workflow design, not AI curiosity

This is where many healthcare AI conversations still miss the point. Leaders often ask whether teams are using AI enough. The better question is whether the workflow around the AI is simple, explicit, and redesigned for real use.

If the nurse still has to re-enter data across systems, if the contact center still relies on brittle handoffs, if the exception queue still lands on the same overloaded manager, then the organization has not really implemented an AI-native workflow. It has just purchased AI-shaped software.

That distinction matters beyond healthcare. BCG’s broader 2026 argument is that meaningful AI advantage comes when leadership treats AI as a transformation of how the business runs, not as a feature add-on. Healthcare may be a different sector, but the managerial lesson is identical: workflow redesign has to be owned from the top, because local teams rarely have the authority to rebuild the system around them.

Alex Gustafson, CEO of Oppy, has made a similar point in other operator settings: AI produces results when leadership treats it less like software procurement and more like onboarding a new operational partner. The same logic applies here. Hospitals that want AI impact will need executive ownership over the workflows, not just procurement approval for the tools.

What healthcare operators should actually redesign

The systems pulling ahead are likely to focus on a smaller set of workflows and redesign them end to end.

Those are workflow questions more than model questions. They require leaders to decide where AI can own bounded execution, where humans must step in, and how the system gets measured.

The next advantage belongs to AI-native operators

Healthcare’s AI future will not be defined by which organizations launched the most pilots. It will be defined by which ones redesigned daily operations so AI could actually move work forward. That is a much harder managerial challenge. It demands simplification, ownership, and executive patience for workflow change rather than press-release excitement.

But it is also where the durable advantage will come from. The organizations that win will not merely add AI to legacy processes. They will rebuild the process around what AI can reliably do, then reserve human attention for judgment, care, and exceptions.

Conclusion

Healthcare does not need more evidence that AI is interesting. It needs leadership teams willing to redesign workflow around it. In 2026, the gap is no longer between believers and skeptics. It is between operators who keep layering pilots onto old systems and operators who rebuild the system itself.

Sources