I keep watching healthcare organizations make the same mistake with cloud migration.
They treat it like an IT project. Move the data, update the infrastructure, declare victory. Six months later, clinicians are quietly working around the new system because it doesn’t match how care actually happens.
The technical migration succeeds. The transformation fails.
This pattern repeats across health systems of every size. The warning signs are always the same. The costs are always invisible until something breaks. And the fix always requires admitting that technology alone never solves organizational problems.
The Misconception That Derails Everything
Healthcare leaders think cloud migration means moving data to AWS or Snowflake and checking a box.
What they miss is that this fundamentally changes how clinical decisions get made.
I watched a mid-sized health system migrate 800 tables to Snowflake. The IT team executed flawlessly. Data moved cleanly, security protocols were solid, performance improved. But three months in, the clinical analytics team was still pulling reports from the legacy system.
The warning signs started subtle.
IT kept getting requests to “just keep the old system running a bit longer.” In steering committee meetings, clinical leaders said things like “the new reports don’t quite capture what we need” but couldn’t articulate why.
What they eventually discovered was years of institutional knowledge baked into the old system. Custom fields tracking things specific to their patient population. Workarounds clinical staff had developed over time. Even the way data was labeled reflected how their clinicians actually thought about care.
The new cloud system was cleaner and more efficient. But it was generic.
Nobody had mapped how a cardiologist’s morning workflow used that data versus how an ER physician needed it structured. By the time they realized this, they’d spent six months and significant budget. Clinicians had quietly gone back to old habits because the new system made their jobs harder.
Research confirms this pattern. 71% of enterprises struggle with moving data into the cloud. In healthcare, where every data point represents a clinical decision affecting a real patient, that struggle becomes a crisis.
What Gets Lost In Migration
The biggest surprise in workflow mapping is always how much critical information flows through informal channels that aren’t in any system.
Organizations discover their “data systems” are just documentation repositories. The real information exchange happens in hallway conversations, quick phone calls, and institutional knowledge that lives entirely in people’s heads.
I’ve seen organizations map what they thought was a straightforward patient handoff process, only to discover nurses making critical safety decisions based on unwritten rules. “I know Dr. Johnson always wants to be called if this particular lab value trends this way, even though it’s technically in range.”
That knowledge is nowhere in the EMR. Nowhere in protocols. It’s just experience that keeps patients safe.
When you migrate systems without capturing that context, you lose the intelligence that actually matters. The technical migration succeeds. Patient safety quietly degrades.
Another consistent discovery is how much workaround behavior exists that leadership never sees. The “integrated system” is actually held together by one person who manually reconciles data every morning before rounds. Or clinicians have created elaborate Excel macros to translate between systems.
I watched one organization discover their impressive medication reconciliation metrics were only accurate because a pharmacist spent three hours daily fixing data quality issues before they hit reports.
Leadership thought they had a great system. They actually had a great pharmacist compensating for a broken system.
The journey mapping also reveals how differently departments use the same data. What the lab calls “specimen collection time” and what the ED calls “specimen collection time” might be two different moments in the process, stored in the same field. Nobody realizes this until workflows are mapped and the data never quite makes sense for operational decisions.
These aren’t technical problems. They’re organizational reality problems that technology alone can’t fix.
The Trust Problem Nobody Talks About
When organizations force adoption by decommissioning legacy systems too quickly, clinicians find workarounds that are often worse than either system.
Shadow IT emerges. Manual processes multiply. Excel spreadsheets with PHI start flying around.
The real cost is invisible until something breaks.
I’ve seen situations where clinicians maintain their own Excel trackers for high-risk patients because they don’t trust the new system’s alerts. Now you have two versions of truth. The official cloud system shows one thing. Actual clinical decision-making happens in spreadsheets that aren’t version-controlled, aren’t backed up properly, and aren’t integrated with anything.
The financial impact compounds quietly.
You’ve paid for cloud infrastructure that’s underutilized. You’re still paying indirect costs of old workflows through staff time. A nurse spending 20 minutes daily manually reconciling between systems. Multiply that across 50 nurses, and you’ve lost the equivalent of a full-time position just to workarounds.
Then there’s the opportunity cost. You can’t build the advanced analytics or predictive models you migrated to cloud for in the first place. Your data is fragmented across official and shadow systems.
But the most expensive cost is clinical risk that doesn’t materialize until it does.
When critical patient information lives in someone’s personal spreadsheet and that person is on vacation, you have gaps in care continuity. I know of a case where a medication reconciliation error happened because the cloud system and the nurse’s manual tracker had different information, and the physician trusted the wrong source.
No patient harm occurred. But it easily could have.
The shadow IT problem tells you that clinicians have so little faith in the official system that they’re willing to risk compliance violations and add work to their day just to avoid using it.
That’s not a technology problem. That’s a trust problem. And those are much more expensive to fix.
The data supports this concern. 83% of IT professionals report their coworkers store company information on unsanctioned platforms. In healthcare, where breach costs average over $10.9 million, this creates vulnerabilities that didn’t exist before migration.
The Coming AI Reckoning
What worries me most is that we’re about to layer AI and predictive analytics on top of these same broken foundations.
Organizations are racing toward machine learning models and clinical decision support tools without first ensuring their underlying data actually reflects clinical reality.
If your data is fragmented, if clinicians don’t trust it, if institutional knowledge isn’t captured, then your AI is just going to automate and scale those problems at machine speed.
I’m seeing health systems invest heavily in predictive models for things like readmission risk or sepsis detection. But these models are only as good as the data quality and clinical adoption underneath them.
If clinicians already have workarounds because they don’t trust the system, they’re certainly not going to trust an AI recommendation based on that same system.
You’ll end up with alert fatigue on steroids. Sophisticated algorithms generating predictions that clinicians ignore because they’ve learned the underlying data isn’t reliable.
The other concern is the growing gap between what technology can do and what healthcare organizations can actually absorb. Cloud infrastructure and AI capabilities are advancing faster than most health systems can change their culture and workflows.
We’re creating this expectation that technology will solve workforce shortages and burnout. But if we implement it the same way we’ve been doing it, top-down and IT-driven without real clinical integration, we’ll just add new burdens on top of existing ones.
What I think we’re setting up for is a reckoning in about three years where organizations realize they’ve spent tens of millions on advanced technology stack that nobody uses effectively.
The winners will be the ones who slow down now, get the fundamentals right, data quality, clinical trust, workflow integration, before they chase the next shiny object.
But the pressure to innovate and the fear of being left behind makes that discipline really hard to maintain.
What Successful Organizations Do Differently
The organizations that got it right treated the first month as discovery, not planning.
They weren’t building project timelines. They were mapping how work actually happens.
I watched one health system spend their entire first month just observing and documenting workflows. Data analysts shadowed clinicians, sat in on care coordination meetings, watched how nurses used the current system during shift changes.
They weren’t asking “what features do you want?” They were documenting “here’s how a discharge decision actually gets made, here’s every data point touched, here’s every system accessed.”
What separated them was they funded this discovery properly.
Most organizations want to jump straight to vendor selection and technical architecture because that feels like progress. The successful ones allocated real budget and time, usually 60 to 90 days, before making any technology decisions.
They created what they called a “clinical data journey map” that showed how information flowed through actual patient care scenarios, not theoretical workflows.
They also identified their informal leaders early. Not just the CMO or department heads, but the charge nurses, the senior physicians that everyone respects, the analysts who really understand the data.
These people became part of the core team from day one, not brought in later for input. And critically, they had veto power.
If a clinical leader said “this approach won’t work in practice,” the project paused until they solved it, even if it delayed timelines.
The other thing they did differently was pilot ruthlessly.
Instead of big-bang migrations, they’d pick one department or one use case, implement it fully, measure actual clinical outcomes, and only then expand. They gave themselves permission to fail small and learn, rather than commit to massive transformation and hope it works.
The key difference was redefining what “success” meant before they started. Not technical milestones like “migrate X tables by Y date” or “achieve Z% uptime.” Those are important, but they’re not success.
Success was measured by whether clinicians were making better, faster decisions that improved patient care.
The Behavioral Indicators That Actually Matter
The honest timeline is about 90 days before you see real signals.
But most organizations want to declare victory at 30 days based on technical metrics that don’t mean anything for actual care delivery. They’ll celebrate “migration complete” or “system uptime at 99.9%” while clinicians are quietly struggling.
The early indicators I watch for are behavioral, not technical.
First, are clinicians logging into the new system without being reminded? If you’re still sending emails saying “remember to use the new dashboard,” that’s a red flag.
Second, are they asking for enhancements rather than workarounds? When a physician says “could we add this field?” instead of “can I just export to Excel and do it myself?” that tells you they see the system as theirs, not IT’s.
Third, look at the questions coming into your help desk.
Early on, you’ll get basic “how do I” questions. If those persist past 60 days, adoption isn’t happening. But if questions shift to “why does this data look different than I expected?” or “can we integrate this other source?” that means they’re actually using it for real clinical work.
The metric that matters most is time-to-insight for clinical decisions.
In the old system, how long did it take a care coordinator to pull together everything needed for a high-risk patient review? If that’s faster in the new system and they’re actually using it that way, you’re succeeding.
If it’s technically faster but they’re still doing it the old way, your transformation is failing regardless of what your project dashboard says.
Real success shows up in changed behavior, not completed tasks.
The First Conversation That Changes Everything
If you’re advising a healthcare organization starting their cloud transformation tomorrow, the first conversation needs to happen before any technology decisions get made.
The room needs to include your CEO or COO, your CIO, your CMO or VP of Clinical Operations, and representatives from frontline clinical staff. Not just department heads. The charge nurses and senior physicians that everyone actually listens to.
The conversation is simple but rarely happens.
Define three specific clinical workflows that must measurably improve. Not just “work differently,” but actually get better. Maybe it’s reducing the time to identify sepsis risk, or improving care coordination for complex patients, or giving physicians back an hour a day.
Then structure the entire transformation around delivering those outcomes, with clinical leaders having equal authority to IT leaders in making decisions.
This means being willing to say no to technically elegant solutions that don’t serve those clinical goals. It means potentially spending more time in discovery and less on implementation. It means measuring adoption and clinical satisfaction as rigorously as you measure system performance.
And critically, it means building in the ability to course-correct.
Keep legacy systems longer than you want. Pilot before full rollout. Have formal checkpoints where clinical leaders can pause the project if outcomes aren’t materializing.
The executives who make this decision, to prioritize clinical outcomes over technical completion, will be the ones who emerge from the three-year reckoning with systems that actually transformed care delivery.
The ones who chase speed and efficiency metrics will have expensive infrastructure that clinicians work around instead of with.
It’s a choice between building technology that serves the mission or building technology that just exists.
What I Wish I’d Understood Earlier
Data systems in healthcare aren’t really about data at all.
They’re about trust, relationships, and how people actually think.
I spent years focused on technical elegance, clean architectures, efficient processes. I thought if we could just get the data structured right and the systems performing well, adoption would follow naturally.
But healthcare data lives in a fundamentally human context.
Every data point represents a clinical decision, often made under pressure, that affects a real patient. When we migrate systems or implement new technology, we’re not just moving information. We’re disrupting relationships, challenging institutional knowledge, and asking clinicians to trust something new when the stakes are someone’s health or life.
The technical work is actually the easy part.
What’s hard is understanding that a nurse’s “inefficient workaround” might be a safety mechanism developed over years of experience. That a physician’s resistance to a new dashboard isn’t stubbornness. It’s because the old system matched how they think about patient care and the new one doesn’t.
That data quality problems are usually organizational problems in disguise.
If I’d understood this earlier, I would have spent less time in architecture meetings and more time shadowing clinical workflows. I would have measured success differently from the start.
I would have recognized that the most important stakeholder in any data transformation isn’t the CIO or the project sponsor.
It’s the frontline clinician at 2 AM making a decision that could save or cost a life.
Everything else is just infrastructure serving that moment. Once you really internalize that, you build systems differently. You lead transformations differently. And you’re a lot more humble about what technology can and can’t do.
The healthcare organizations that understand this will be the ones still standing when the reckoning comes. The ones that don’t will have expensive cloud infrastructure and clinicians who’ve learned to work around it.
The choice is being made right now, in the first conversations about cloud migration, before any code gets written or any data gets moved.
Make it count.
