Fixing Documentation Gaps with Healthcare Workflow Automation: Where Hospital AI Goes Next

Hospitals have moved from 200 tons of paper charts to agentic AI in about six decades, but according to William A. Conrad, MD, Director of Utilization and Physician Advisor at Providence, the technology itself was never the hard part. The hard part is figuring out where AI workflow automation actually removes friction, and where it needs firm boundaries.

That was the throughline of a recent Becker’s Healthcare webinar, "From Paper Charts to Orchestrated Departments: Where Hospital AI Goes Next," featuring Dr. Conrad and Sarah Galyon, Senior Director of Healthcare Solutions at Intellistack. Galyon spent 17 years in clinical process design at The University of Tennessee Medical Center before joining Intellistack, and the two walked through decades of healthcare technology history before landing on where hospitals should focus right now: small, high-friction workflows that are ready for automation today, and the governance structures that need to exist before AI goes any further.

Two moments from the conversation stood out. One is a concrete example of AI already saving a health system real time and real revenue. The other is a warning about what happens if the industry doesn’t get ahead of AI’s expansion into clinical work.

The Easiest Win: Fixing Clinical Documentation Integrity with Healthcare Process Automation

Dr. Conrad calls clinical documentation integrity, or CDI, "the lowest hanging fruit" his team has found for AI. The problem is simple to describe and expensive to ignore: a physician documents a diagnosis like pneumonia, but the medical record actually supports a more complex condition, such as pneumonia with acute respiratory failure and sepsis. That gap between documentation and clinical reality directly affects reimbursement.

The traditional fix is a manual, nurse-driven query and document approval workflow, where CDI teams flag discrepancies and send a request to the physician’s inbox asking them to amend the record. When physicians don’t respond, someone has to chase them down by phone, email, or in the hallway. He describes it bluntly: "It’s a very, very inefficient system."

To that end, working with Intellistack, a major health system has built an automated escalation system that pulls open queries and reaches physicians directly by text and email, removing the manual chase entirely. The results were immediate. In the first ten weeks, time in queue dropped 23%, with average delinquency falling from 2.5 days to 1.2 days. Most physicians now resolve a query after a single notification.

Dr. Conrad and Sarah Galyon on building the automated escalation system and the 23% drop in query time.

At one hospital alone, CDI queries represent roughly $400,000 a month, or close to $5 million a year, in accurately captured care. Speeding up that process isn’t a minor efficiency gain. As Galyon put it, this kind of automation isn’t about replacing anyone. It’s about "decompressing your work burden," so physicians and administrators can spend less time on manual follow-up and more time on patient care.

The Harder Question: Who Governs the AI?

The second clip shifts from a working example to a caution. Dr. Conrad, who spent a decade as an examiner for the American Board of Emergency Medicine, points out that physicians have always been held to a rigorous, standardized bar of competency and consistency. AI, so far, hasn’t been.

"We have to put our arms around it and say, yes, this is going to be part of what we do," he explained, "but in putting our arms around it, we have to guide its direction." He was direct about what that means in practice: oversight and regulation are necessary to ensure AI is both accurate and applied consistently, the same way healthcare delivery itself is expected to be consistent across the country.

Dr. Conrad on the need for oversight, certification, and consistent regulation as AI expands in healthcare.

It’s a notable position from someone actively deploying AI inside his own health system. Dr. Conrad isn’t arguing against adoption. He’s arguing that adoption without guardrails is how healthcare ends up accountable to a tool nobody formally vetted. Strong enterprise AI workflow governance, with compliance workflow automation built in, is how health systems avoid that outcome.

The Takeaway

These two moments capture the same underlying idea from opposite angles. CDI automation shows what’s possible when a health system identifies a specific, well-scoped pain point and applies AI to it deliberately. The governance discussion is a reminder that the same discipline needs to extend to how AI is evaluated and regulated as its role grows from single-patient tasks to full department workflow orchestration.

For health systems still figuring out where to start, Dr. Conrad’s advice is straightforward: find your own version of CDI, a narrow, high-friction process with a clear cost that is ready for healthcare workflow automation, and build from there.

To see how IntelliStack is helping healthcare teams orchestrate and automate workflows like this one, visit IntelliStack.com/solutions/industry/healthcare. Or, watch the full webinar on demand.

‍