The Hardest Part of Healthcare AI Transformation Isn’t the Technology, It’s Change

AI can transform healthcare workflows, but only if we prepare people to work differently.

Artificial intelligence is moving at warp speed. The technology is advancing faster than most organizations can evaluate it, govern it, implement it, and train their workforce to use it.

Healthcare leaders are being introduced to new AI capabilities almost daily: generative AI, copilots, ambient documentation, predictive analytics, autonomous agents, AI-assisted coding, revenue cycle automation, regulatory intelligence, clinical decision support, and more.

It is tempting to think the primary challenge is choosing the right technology. I don't believe it is. One of the greatest challenges of AI transformation will be preparing organizations, and the people within them, to work differently. That is why healthcare AI change management must be treated as a leadership priority. Technology can be purchased, but transformation has to be led.

AI Implementation Is Not AI Transformation

There is an important distinction between implementing an AI tool and transforming a workflow.

An organization can purchase an AI solution, integrate it into its technology environment, provide employees access, and call the implementation complete. But leaders still need to ask what actually changed. Did the workflow improve, were unnecessary steps eliminated, and did roles or decision-making change? Did employees gain meaningful capacity, did quality improve, and were new risks identified and controlled?

If the organization simply inserts AI into an inefficient process, it may end up automating inefficiency rather than transforming the work. That is why I believe AI transformation must begin with the workflow—not the tool. Leaders must understand how the work happens today and then determine how it should happen tomorrow.

People Need to Understand What AI Means for Their Jobs

We cannot talk about AI transformation without acknowledging the concern many employees have: What does this mean for me? Employees may wonder whether AI will replace their jobs, change their responsibilities, alter how performance is measured, or make decisions about their work. They may also question whether they are expected to trust something they do not completely understand and whether they will be responsible if the AI is wrong.

These are legitimate questions. If leadership doesn't address them, employees will answer them for themselves—and fear fills information gaps quickly.

Healthcare organizations need to communicate clearly about the role AI is expected to play.

In many workflows, the goal should not be to replace expertise. It should be to remove friction around expertise.

Remove the Work Around the Work

Think about how highly trained healthcare professionals spend their time.

Compliance professionals search multiple regulatory websites and payer policies, auditors manually organize documentation, coders search for guidance, and physicians spend enormous amounts of time documenting care. Revenue cycle professionals research denials, while leaders compile information from multiple systems before they can make decisions. Much of this is what I call the work around the work. It is necessary, but it isn't always where human expertise creates the greatest value. AI has the potential to absorb portions of that burden.

Imagine a compliance professional beginning the morning with regulatory changes already identified, sourced, summarized, classified, and prioritized for review. The professional's time shifts from finding the information to determining what the information means.

That is a very different way of working, and it is where AI begins to augment expertise rather than simply automate tasks.

Roles Will Change

Healthcare organizations should expect AI to change job responsibilities. That does not automatically mean jobs disappear. In many cases, the composition of the job changes. An auditor may spend less time assembling information and more time evaluating exceptions, while a compliance professional may spend less time searching for regulations and more time interpreting risk. A coder may increasingly validate AI-assisted recommendations rather than manually constructing every code combination. A revenue cycle analyst may focus less on identifying anomalies and more on determining their causes, while a manager may spend less time compiling reports and more time making decisions based on them.

That creates a new leadership responsibility: We need to redesign roles around where human expertise creates the greatest value.

AI Literacy Must Become a Workforce Competency

Employees cannot responsibly use technology they do not understand. That does not mean everyone needs to become an AI engineer. But healthcare professionals increasingly need a practical understanding of AI.

They should understand what LLMs and AI agents are, why hallucinations occur, and why AI confidence does not necessarily equal accuracy. They also need to understand source validation, what data should and should not be entered into AI systems, when an output requires validation, when human escalation is required, and how the organization's AI governance applies to their work.

For leaders, the competency level needs to go further. Executives, compliance officers, privacy officers, clinical leaders, revenue cycle leaders, and board members need enough AI literacy to govern what they are being asked to approve. AI literacy is quickly becoming part of leadership literacy.

Domain Experts Need to Help Design the AI

This may be one of the most important elements of successful healthcare AI transformation.

The people who understand the work need to participate in redesigning the work.

IT cannot independently determine how an AI coding workflow should operate. A vendor should not independently establish when an AI-generated compliance recommendation is reliable enough to act upon, and a data scientist should not independently determine what constitutes a material regulatory risk. Those decisions require domain expertise.

Depending upon the use case, healthcare AI requires cross-functional collaboration among technology, operations, compliance, privacy, security, legal, clinical expertise, revenue cycle, and data teams.

This is where distributed AI literacy becomes so important.

Domain experts do not need to build the technology, but they absolutely need enough understanding to help design, test, challenge, and govern it.

Don't Ask Employees to Trust AI Blindly

"Trust the AI" should never be the implementation strategy. Trust should be earned. Employees need to understand where information came from, what the AI is and is not designed to do, how outputs are validated, and what known limitations exist. They should also know what happens when the system is uncertain, how they can challenge an output, how errors are reported, and who is accountable.

This is particularly important in healthcare because experienced professionals will sometimes know that an AI recommendation is wrong. We need them to speak up. A strong AI culture should encourage employees to challenge AI, not discourage them from doing so.

Human Overrides Are Valuable Information

Suppose an experienced auditor repeatedly overrides an AI recommendation. Leadership should not immediately conclude that the employee is resistant to AI. It should ask why. Perhaps the AI is missing clinical context, the underlying knowledge base is incomplete, or the risk threshold is wrong. The model may not understand a payer-specific exception, or the human may be identifying something the system was never trained or designed to recognize.

Those disagreements can become extremely valuable feedback. Human expertise should help improve the AI system—not be silenced by it. This is another reason I believe in Human-in-Governance™. Humans do not merely approve AI outputs. They help establish, monitor, challenge, and continuously improve the environment in which AI operates.

Leadership Must Create Psychological Permission to Learn

AI is developing so quickly that organizations cannot expect employees to master it before they begin using it.

There will be experimentation, questions, mistakes, and things we try that do not work. That requires a different leadership mindset.

Employees need appropriate, governed environments where they can learn.

Leaders should encourage employees to ask whether AI could improve a workflow, what would happen if a step were redesigned, and which manual activities add little value. Teams should also identify where human judgment is essential, what could go wrong, and how they would know.

Curiosity should be encouraged within clear governance boundaries. That is very different from either extreme: "Use AI everywhere" or "Don't use AI because it is risky." Neither is a strategy.

Change Leadership Requires Transparency

If AI will materially affect how employees work, leadership should communicate early and often.

People should understand why the organization is adopting AI, what problem it is trying to solve, and what will and will not change. Leadership should explain what remains human-owned, which new skills employees will need, how success will be measured, how concerns will be addressed, and what happens when AI gets something wrong.

Transparency is particularly important when AI affects productivity measurement, staffing, performance management, clinical decisions, reimbursement, or other consequential activities.

Employees should not discover significant AI-driven workflow changes after the technology has already been deployed.

Start Small, but Design for Scale

Organizations do not need to transform everything at once. In fact, they shouldn't.

Select a workflow where the problem is meaningful, the outcome can be measured, and the risk can be appropriately governed.

They should map the current state, establish a baseline, design the future state, and define AI and human responsibilities. They should identify risks, build controls, train the people involved, and then pilot, measure, listen, and adjust before determining whether the workflow is ready to scale.

But even when starting small, leadership should think ahead.

Leadership should consider what happens when one AI agent becomes ten, when AI moves from summarizing information to taking actions, when multiple departments use the same technology differently, or when an AI-enabled process becomes business-critical.

The operating model needs to mature with the technology.

Accountability Must Change With the Workflow

When work changes, accountability must be reconsidered. If AI performs a task previously completed by an employee, the organization must determine who now owns quality. It also needs to establish who investigates an incorrectly routed case, who owns a human-approved AI recommendation, who evaluates the impact of a vendor model change, and who has authority to stop the system when it begins producing unusual results. These questions cannot be answered with "The AI did it." AI can perform work, but it cannot absorb organizational accountability. That remains human.

Measure Adoption Differently

Organizations should also be careful about how they measure AI adoption.

More usage does not necessarily mean better adoption. If employees are using AI frequently but incorrectly, bypassing controls because the workflow is cumbersome, or saving time while increasing rework, the implementation is not successful.

Better measures include appropriate use, output quality, human override rates, reductions in low-value work, and time redirected toward higher-value activities. Organizations should also assess employee confidence and competency, compliance with governance requirements, and measurable workflow improvement. The goal is not maximum AI usage. The goal is effective human-AI collaboration.

The Organizations That Learn Fastest Will Have an Advantage

AI will continue to evolve. The tools organizations implement today may look very different twelve months from now. That means the long-term competitive advantage may not be a particular AI platform. It may be the organization's ability to continuously adapt.

The real test is whether leaders can evaluate new capabilities quickly, domain experts can participate intelligently, and governance can adapt without starting over. Employees must be able to learn new ways of working, workflows must be able to evolve, and the organization must recognize when something is not working and change course.

That organizational capability may ultimately matter more than any single technology decision.

The ProCode Perspective

For more than four decades, I have watched healthcare navigate major transformations in technology, regulations, payment models, and workflows. But one lesson continues to hold true: Technology alone does not transform an organization. Leadership does.

AI may be one of the most powerful technologies healthcare has ever encountered, but the organizations that succeed will not simply be those that purchase the best AI. They will be the organizations that prepare their people to work with it.

That means building AI literacy, redesigning workflows, empowering domain experts, creating governance, defining accountability, listening to employees, measuring outcomes, learning continuously, and preserving human authority where it matters.

AI transformation is not ultimately about teaching people how to use AI. It is about redesigning how people, technology, and intelligence work together. That is why the hardest part of healthcare AI transformation may not be the technology at all. It will be leading the change.