Streamlining Healthcare IT Support with Machine Learning: A Real-World Win

In healthcare, technical inefficiencies don’t just cost time—they disrupt care. When clinicians can’t access systems or tools they need, it delays treatment and increases frustration. At Cactus Healthcare Resources, we focus on solving problems that actually matter—even if they’re not flashy. Some of the best applications of data science aren’t the most complex—they’re the most useful.

The Problem: Misdirected IT Tickets Were Wasting Days

One of our clients, a large health system, was struggling with a familiar issue: IT support tickets were being routed manually—and incorrectly. A user would submit a ticket, and a generalist service desk agent had to guess which of 30+ IT teams should handle it.

The impact was real:

• Roughly 50% of tickets were initially routed to the wrong team
• Each incorrect handoff added up to a day in delays
• Tickets regularly bounced between 3–5 teams before reaching the right destination

Even straightforward problems that could’ve been resolved in hours were taking 2–3 days. That’s an eternity when you’re dealing with system outages in EHRs, billing, or patient portals.

The Fix: A Simple, Custom Machine Learning Model

Instead of pitching a bloated vendor platform or expensive off-the-shelf AI, we did what should be standard: we worked with what the client already had.

Using their existing ticketing system and infrastructure, we:

• Analyzed 350,000 historical tickets
• Built a natural language processing (NLP) model to classify ticket descriptions
• Exposed the model via a lightweight API integrated directly into their helpdesk system
• Automatically routed tickets when the model was confident, while maintaining human review where it wasn’t

The model learned that terms like “charge router” or “transaction type” pointed to the Hospital Billing team, while phrases like “Cadence schedule freeze” were Epic-specific, and even further distinguishable by module.

No hype. No layers of abstraction. Just clear, simple patterns identified and used to make better decisions.

The Results: Quietly Transformative

• 81% routing accuracy (up from ~50%)
• 2–3 days faster resolution times on average
• Up to 60% faster resolution for specific teams
• Reduced manual workload for service desk staff
• Immediate improvements in user satisfaction

And when the model wasn’t confident? The fallback was the old manual process—no worse than before. That made this not just effective, but low-risk.

Why It Works

This wasn’t a moonshot. It was practical, and it paid off because it was:

• Inexpensive – cloud cost is pennies per hour
• Non-disruptive – no new systems to buy or learn
• Sustainable – retraining needed just once or twice a year
• Safe – worst-case performance matches the baseline

The simplicity is the strength. And that’s a theme we see again and again.

Beyond This Project: What Real Data Science Looks Like

This is what we do at Cactus: solve the real problems. Not the ones in case studies or startup decks—but the ones your staff deal with every day.

We’ve seen that some of the highest ROI projects in healthcare aren’t the most complex. They’re often the ones that:

• Use existing data
• Avoid major system overhauls
• Are built for how people actually work

Whether it’s ticket routing, forecasting workloads, or optimizing resources across departments, our team builds solutions that are fast to deploy and grounded in reality.

Let’s Talk

If your IT teams are bogged down by routing inefficiencies, manual triage, or just plain noise, let’s have a conversation. We’ll help you figure out if a similar approach can help—and if not, we’ll tell you that too.