A conversation with Crystal Broj, Enterprise Chief Digital Transformation Officer at MUSC

Just about everyone has an AI story right now. Some are impressive. Some are cautionary tales. Most of the messy ones in the middle never get talked about publicly.

That’s exactly why we sat down with Crystal Broj, Enterprise Chief Digital Transformation Officer at the Medical University of South Carolina (MUSC). In four years, she’s grown her digital transformation team from one person to nearly ten, and she’s led dozens of AI and automation initiatives across patient access, revenue cycle, and clinical operations.

Wherever you and your organization might be in your AI journey, her learnings offer some insight in how to structure your next roll-out, what to watch out for, and even when to call it quits.

Why AI Feels So Polarizing in Healthcare Right Now

Pick any week and you’ll find two kinds of AI headlines: the, “we saved thousands of hours,” success stories and the, “we pulled it out after three months,” failures and both are happening at the same time. Where one system sees immense savings and improvement, another finds waste. Those differing journeys are making it hard for healthcare teams to know what to trust and what to implement.

Crystal’s take: the confusion isn’t really about AI. It’s about expectation versus reality.

The important thing to remember is that AI in Healthcare will amplify whatever workflow it touches. If your process has unnecessary steps, automation just gets you to those unnecessary steps faster. Crystal summarized it well, “It’s just going to do what you’re doing, but faster. So if you have hiccups, you’re going to get to your hiccups faster.”

Document the workflow, fix it, then automate.

Part 1: The Technology Usually Isn’t the Hard Part

MUSC automated mammogram outreach and rolled it out to their upstate South Carolina location. Within one night, 100 people scheduled appointments. A win, right?

Then Crystal got a call: “Whatever you did, turn it off.”

The slots were filling up faster than the staff could handle the downstream work. They didn’t have enough people for that volume.

The technology did what it said it would do, however, the process wasn’t ready for it.

This is the lesson Crystal says she learned early, and still sees organizations miss: technology is the easy part. People and process are where implementations live or die.

Part 2: Change Management Is Not a Checkbox

MUSC rolled out a pre-registration tool iteratively, starting simple (confirm or cancel your appointment), then adding features with each wave. Good adoption across new offices. But at the original offices? Usage was dropping.

When they dug in, they found front desk staff telling patients, “Don’t use that thing. Just come see me.” Why? Because there had been early bugs, those bugs had been fixed, but nobody had gone back to tell the people at the beginning of the line.

The tool had improved. The staff didn’t know, and already become comfortable with their workarounds since healthcare doesn’t stop for technology bugs.

Crystal’s hard-won rule: ongoing training and reinforcement aren’t launch tasks, they’re ongoing operations. Staff turns over. Systems get updated. If the people touching the workflow every day don’t understand what the tool does and why it matters, adoption stalls no matter how good the technology is.

And beyond training, there’s a deeper communication job: making sure people understand that AI isn’t coming for their jobs. It’s coming for their clipboard.

Crystal goes on to explain, “We still need you here. You just might have more time to actually say, ‘Hey, how’s it going?’ or help someone navigate a referral they’re nervous about.”

AI handles the busy work, people do the human work. Ideally, freeing staff from the busy work allows them to connect with patients more.

Part 3: Prepare Before You Pilot (and Define Success Before You Start)

When stakeholders come to Crystal’s team with a problem they want to solve, the process starts long before any vendor demo. Demos get people excited in ways that can derail the actual goal. The excitement is a good sign, but there’s a few things that need to be defined before planning a pilot or proof of concept.

Her preparation checklist:

1. Start with the problem, not the tool. If there’s no clear problem statement, there’s no project.
2. Map the workflow honestly. Not the SOP version, the real version with every workaround and detail. Why does this person do step two differently than everyone else? Are they working around a gap? That gap matters.
3. Identify the bottlenecks. The step that takes the longest is usually where automation delivers the most volume, or friction.
4. Align stakeholders before the demos. Know what you actually need before you see what a platform can do.
5. Build the dashboard before you launch. Crystal’s mantra: no project goes out without a dashboard. You need to know what you’re measuring, what baseline you’re starting from, and what success looks like before you begin, not after.

That last point is key when having a conversation with leadership when renewal time comes. “It’s great,” isn’t enough. “We hit a 3:1 ROI, and here’s the data to show the transformation,” makes the decision to renew obvious.

Crystal distinguishes between hard green dollars (direct cost savings, time recovered) and soft green dollars (care quality improvements, patient satisfaction, closed care gaps). Both matter. Both belong in the story you tell.

Part 4: Don’t Forget About Governance

Governance tends to get framed as the thing that slows everything down. Crystal pushes back on that hard.

Without governance, you get shadow IT, small teams running their own tools that don’t connect to anything else. There are redundant platforms, possibly ones that don’t have a BAA in place. You might get a doctor adding questions to a shared patient questionnaire without anyone knowing how that affects everyone else using it.

Good governance answers a few fundamental questions: Who approves new initiatives? Who owns the outcomes? How do we measure value? And who gets to change things after launch?

Governance also includes keeping an eye on your broader tech stack. If a point solution you adopted two years ago hasn’t kept pace with your core systems, that’s a governance conversation that allows for a better fit to be brought in.

Part 5: Stopping a Project Isn’t Failure

This might be Crystal’s most important point, and key to implementing most new technologies, not just AI.

When a pilot isn’t working, there’s enormous pressure to keep going, especially when someone championed the tool and staked some of their reputation on it. That sunk cost thinking is exactly what leads organizations to double down on bad implementations. A technology that doesn’t fit only snowballs losses later.

Crystal recently had a team come to her and say, nervously, “This isn’t working out.” Her response: “Great.”

They were surprised. She wasn’t.

“If you found that it didn’t save time, and it was more effort than we thought, did we learn something? Then we succeeded. That is exactly what a proof of concept is for.”

She uses three questions to guide the pivot decision:

• Is the problem still real?
• Is this still the right tool?
• Are we willing to change the process?

If the answers push toward no, walking away creates space for a better solution. And in AI right now, there’s almost always a better solution in development.

The Takeaway

Crystal’s closing advice to anyone starting, or stuck, in their AI journey:

“Start small, but be intentional. Fix your workflows before you add AI. Invest as much in communication, training, and change management as you do in the technology — probably more. Focus on outcomes, not optics. And remember: sustainable AI is built, not bought.”

AI is powerful. But it’s human-led. The technology can only go as far as the people and processes behind it.

Watch the full webinar, Prepare, Pilot, Pivot: Navigating the AI Journey in Healthcare, featuring Crystal Broj, Enterprise Chief Digital Transformation Officer at MUSC.