Electronic Health Reporter

By Gilda D’Incerti, CEO and Founder, PQE Group

The healthcare industry has spent the better part of two decades digitizing itself, and it has plenty of evidence to show for it. Electronic health records replaced paper charts. Revenue cycle systems automated financial processes. Telehealth moved portions of care delivery onto screens. Interoperability initiatives attempted to connect information across systems. Cloud computing, analytics, and mobile applications followed.

The scale of that transformation is difficult to dispute. By 2024, more than 99% of U.S. non-federal acute care hospitals had adopted a certified electronic health record, while 91% of office-based physicians reported using one.

Yet many of the problems those technologies were expected to solve remain stubbornly familiar.

Clinicians still spend too much time documenting. Administrative processes remain fragmented. Patient information can still be difficult to find and interpret. Systems intended to simplify work can introduce their own complexity. Healthcare organizations have accumulated enormous amounts of technology without necessarily changing the underlying processes that determine how work gets done.

That history matters as healthcare moves aggressively into artificial intelligence.

The industry is approaching AI with many of the same assumptions that accompanied previous technology transformations: identify promising applications, select the right technology, implement it, train employees, and measure adoption. Those steps are necessary, but they do not constitute transformation. Healthcare has learned how to implement technology remarkably well. It has been less successful at using technology as a reason to reconsider how work should be performed fundamentally.

That could determine whether AI delivers significant change or becomes another layer of technology on top of an already complicated system.

Healthcare has already made this mistake

The electronic health record is the clearest example.

The promise of digitization was never to replace paper charts with electronic ones. The larger opportunity was to make information more accessible, improve coordination, and create a foundation for better care delivery. Much of that happened. EHR adoption is now nearly universal across hospitals and widespread among physicians, giving healthcare an extraordinary digital infrastructure that did not exist a generation ago.

But digitization did not eliminate the underlying problems of healthcare administration and clinical workflow. In many cases, organizations adapted existing processes to the capabilities and limitations of the software they purchased. Documentation requirements moved onto screens. Information became digital without necessarily becoming easier to interpret. Different systems could exchange information without eliminating the organizational boundaries that separated the people using it.

The lesson is not that EHRs failed. They did not. The lesson is that technology alone cannot transform a process that nobody has taken the time to redesign – that matters as organizations evaluate AI.

Healthcare does not need another generation of technology that makes existing processes faster. It needs to determine which processes should exist in the first place.

AI can make a bad process more efficient

There is an understandable enthusiasm around AI’s ability to automate administrative work, summarize information, process documents, support decision-making, and extract useful information from data that has historically been difficult to use. Those capabilities have genuine potential, particularly in an industry that generates enormous volumes of structured and unstructured information.

The danger is assuming that a technically successful AI deployment is automatically organizationally successful.

Consider a referral process that requires multiple handoffs, redundant documentation, and manual review. An organization could introduce AI to process the information faster and route referrals more efficiently. If the underlying process remains unnecessarily complicated, however, the organization has not really solved the problem. It has simply accelerated a poorly designed process.

The same issue can appear in revenue cycle operations, patient communications, clinical documentation, and virtually any other area in which healthcare organizations have accumulated layers of procedures over time. AI can reduce the human effort required to execute those procedures, but it cannot determine whether the procedures still make sense.

That decision belongs to the organization.

That is why some of the most important questions about AI deployment have little to do with an individual model’s capabilities. They concern workflow design, data quality, governance, accountability, and leadership’s willingness to change processes that have existed for years.

The technology is changing faster than the organization

AI also creates a challenge that healthcare has not encountered at this pace in previous technology cycles. The technology itself is evolving continuously.

Traditional healthcare technology implementations regularly involve lengthy planning, procurement, configuration, validation, and training cycles. AI systems can change significantly during the period an organization evaluates or deploys them. Models are updated, new capabilities emerge, vendors change their products, and organizations discover new applications once employees begin using the technology.

That creates a problem for an industry accustomed to treating technology as a project with a defined beginning and end.

AI is better understood as an ongoing capability that requires persistent oversight. For healthcare and life sciences organizations, that makes validation, governance, and risk management particularly important. The answer cannot be to slow every AI initiative until the technology stops changing. It will not. 

Organizations instead need processes that can evaluate, monitor, and govern technology as it evolves.

That also means treating validation as something more substantive than a final checkpoint before deployment. If the technology, data, workflow, or intended use changes, the organization needs to understand what else has changed.

The technical capability may be new, but the underlying management challenge is familiar. Healthcare has repeatedly underestimated the organizational work required to make technology useful.

The harder question is what should disappear

The most useful AI conversations may therefore begin somewhere other than the technology itself.

Instead of asking where AI can be added to an existing workflow, healthcare leaders should ask what the workflow would look like if they were designing it from scratch with today’s technology.

That question can produce uncomfortable answers.

If AI can review and summarize large volumes of information, should employees continue spending hours performing the same manual review? If systems can extract information from documents, should staff keep entering it into multiple applications? If AI is able to handle routine patient communications, should organizations add an AI tool to an existing call center structure, or should they reconsider how patient communication is organized altogether?

These questions require more than technology expertise. They require operational leaders to examine who owns a process, where decisions are made, why certain steps exist, and what work can reasonably be eliminated.

That is where many digital transformation efforts become difficult. Replacing a system is quite easy compared with persuading multiple departments to abandon processes they have spent years building around that system.

Healthcare leaders should also consider a more uncomfortable possibility: AI could become a way to avoid organizational changes that have been difficult for years.

Redesigning a referral process may require departments to surrender control. Deleting redundant documentation may mean changing responsibilities. Consolidating administrative functions can affect budgets, reporting structures, and jobs. Those decisions are difficult, particularly in large healthcare organizations where every process has accumulated stakeholders over time. Buying an AI system can feel considerably easier.

That creates a real risk. Instead of asking whether a process should exist, organizations may ask how cheaply AI can perform it. Instead of eliminating unnecessary administrative work, they may automate it. Instead of simplifying the organization, they may use technology to make its complexity more efficient.

That is preservation through automation, not transformation.

The most consequential AI decisions may therefore involve work technology should make unnecessary, rather than simply work it can perform. Healthcare leaders will have to decide what processes, handoffs, documentation requirements, and administrative tasks should disappear rather than assuming every existing responsibility deserves an AI-enabled version.

That conversation will be considerably harder than choosing an AI vendor. It may also be considerably more important.

AI could expose the weaknesses of everything that came before it

There is a tendency to describe artificial intelligence as the next phase of healthcare’s digital transformation. A more useful way to think about it may be as a test of the transformation that has already taken place.

AI depends on the quality and accessibility of the information around it. It depends on reliable workflows. It depends on clear accountability. It depends on people understanding when to trust its output and when to challenge it. It depends on organizations knowing what they are trying to accomplish before they automate the work required to accomplish it.

Those dependencies bring healthcare’s existing weaknesses directly into the AI conversation.

Disjointed data becomes an AI problem. Poorly defined processes become an AI problem. Weak governance becomes an AI problem. Unclear ownership becomes an AI problem. Organizational resistance to change becomes an AI problem when employees are asked to incorporate quickly evolving technology into workflows that were never designed for it.

That is not necessarily a reason for pessimism. It may be one of the most valuable things AI does for healthcare.

Previous technology transformations often stopped short of genuine process redesign. Organizations could implement a new system while the underlying workflow remained largely intact. AI may make that approach increasingly difficult because its greatest value often comes not from performing an existing task more quickly, but from changing how the task is performed.

Healthcare leaders should seize that opportunity rather than repeat past implementation patterns.

The question should not simply be where AI can be deployed. It should be what healthcare would design differently if it were not constrained by the processes, organizational structures, and assumptions it inherited from an earlier era of technology.

Healthcare has spent years making technology fit the way it already works. The opportunity with AI is to reverse that relationship and begin redesigning the work around what the technology can responsibly make possible.

That requires more than another successful implementation.

It requires healthcare to finally treat technology as a catalyst for changing the work itself.

Because healthcare has never really lacked technology, it has lacked the willingness to change how it uses it.