Written by Devon Mobley Chief Growth Officer at Calvient

As a leader in healthcare, it's easy to feel pressure to keep up with the times. Technology advances very quickly, and this feels doubly true in the age of AI.

Whether you're ready or not, you feel thrust into a position to think about Artificial Intelligence. Maybe you hoped this day would come after you retire, or perhaps you thought this was initially just another hype cycle that would fade away quickly. It's scary and could do real damage to our industry.

But what I'd like to present to you here is a more optimistic or at least pragmatic picture. One that helps you realize that the issues that have made healthcare technology as much of a burden to care (as a blessing) are much easier to solve with AI. And also perhaps to help you peer inside the black box that is AI and Large Language Models (LLMs).

The reality is, that healthcare is an overly burdened industry. So I propose strategies that would help us relieve burdens on personnel, rather than replace them. In other words, let's find the right way to align this seemingly unstoppable wave of technology with everyone from the front-line staff up to the boardrooms of healthcare.

iPhone Moment

You can take a second to appreciate the trends we've seen even in just the past 25 years. From the advent of the EHR, cloud, mobile, social, VR/AR, blockchain, and now AI. Some of these have had more impact than others. There are iPhone moments and there are Oculus moments. They both have an impact, but the iPhone revolutionized so much about technology in a way that VR/AR never could.

Turns out, we're in an iPhone moment. The way our primary tools of productivity and business—which is to say our software tools/systems—are fundamentally changing before our very eyes. Like when desktop apps went mobile.

Having witnessed the evolution from EHRs to cloud computing, data analytics, mobile technology, and now AI, I can confidently say that AI represents a fundamental change in how we use technology.

Sure healthcare will take some time to catch up. But I think this time it will happen more quickly than we realize. It's important to know and understand your options now, so that when strategy can meet execution, you are prepared.

So first let me start with why AI can be great and then discuss how to think about using it practically in your work.

The Thread Through Healthcare Technology Problems

The reason healthcare information systems are so tedious and difficult to use (read: THEY SUCK!) is because of the following two reasons:

1. Healthcare information is largely "unstructured" (document-based), which is notoriously hard for computers to work with. We've tried to make it work for over two decades, however, with marginal success. Square peg, round hole problem.

2. Healthcare has extremely high administrative demands due to insurance. This requires secure communication, sharing of aforementioned documents (lots of them), and coordination of multiple parties.

The implication is that most tasks feel like it's easier to do manually on paper, rather than in a computer system. The biggest beneficiaries are admins, who can now aggregate the data, etc.

This is because computer software likes working with bits of data, things like numbers. This is partially why EHRs and existing technology feel like something's missing. There's tons of unspoken rules, inferential information, and context that we convey to one another through language and visuals. This context likely makes up over half of the data within healthcare. And computers can't act on it.

AI models such as ChatGPT and Claude however love language (text) and visuals (images). Since we also have a lot ofgreat techniques nowadays for turning documents into text or images (most systems can do this reasonably well) and since AI models now have basic reasoning capabilities, you have a more robust tool for accomplishing healthcare-related tasks than we've ever seen.

Ok, I see the point, but how?

I think one of the ways that AI researchers went wrong is that they were itching to take on clinical tasks. Things that doctor's brains do well (for example) like thinking through diagnoses, working with patients for treatment, etc.

To be sure there are many tedious workflows providers, like creating notes that have a high success rate on claims. It's a part of the business.

But there is so much other work in healthcare that can be made more efficient through AI, such that we free up those moments providers get with their patients. And by starting elsewhere and gaining the trust of clinical and non-clinical staff alike, AI has a bigger chance for success as a transformational tool.

So let me give you what I see as the prime ways to identify where AI can help.

3 Questions to Identify Where AI Can Help

1.) Where is staff training and retention tricky?

The rules e.g. creating a proper prior authorization or how various outpatient teams handle outbound referrals/transitions of care are unique to every healthcare organization. They have their preferences, community relationships, and rapport with payors. What this means is that certain staff have to learn the rules over time. These staff are invaluable. If you spend any time in healthcare you know these people and they are rock stars.

However, this puts a major bottleneck on growth, for the organization or even the individual. The buck in many ways stops with this person, so a well-deserved promotion would mean taking them out of doing the work it seems only they can do with 90+% success.

Why AI Can Help

• Knowledge Base Creation: AI is good at creating knowledge bases that make it easier to extract what you need. Just drop in documents, and images, or even talk it out to a microphone.

• Google on Steroids: Tools like Perplexity are "answer machines" not unlike Google (but way better). As such, these models can be "primed" with the complex rules specific to your organization. The kicker is they are mucheasier to retrieve useful information from. All you have to do is ask a few questions in natural language, instead of searching and clicking around a bunch of documents.

• Leveraging Expert Knowledge: Putting your best staff's brains into these knowledge bases is like having an extra senior employee assisting with training a host of new team members. The value of retaining a central, language-driven knowledge base can help alleviate staff turnover and efficiency issues in a major way.

2.) Where are tasks sitting on hold for a long time?

Tasks such as prior authorizations, referrals, and medication refills have to be coordinated between payors, providers (referring and internal), and patients. A lot of time these communications are sitting in an inbox somewhere, waiting for a response or to be processed by eligibility, billing, or a clinician.

However, it's also really hard to create "if-this-then-that" rules in an EHR or inbox (if you even have that capability). A lot of it has to be reviewed with a decently keen eye for patient safety reasons. Otherwise, you could be filling a now contra-indicated medication or failing to get a payor more records (delaying care).

Why AI Can Help

• Agent Functionality: When prompted correctly, AI models can act as "agents" who have one, singular job. They can read documents or messages and provide a context-sensitive summary/action plan with little effort.

• Rapid Summarization: AI is good at ingesting, contextualizing, and summarizing large volumes of data. Muchbetter than humans. Even simple tools like a suggested next action can go a long way in speeding workflows up.

• Workflow Acceleration: Most of these applications of AI agents act as "workflow accelerants" - so where automation is hard or impossible, they can at least help the humans working (and reduce errors in the process).

3.) Where Do We Rely on Reactive Data Analysis Instead of Proactive Insight?

Many healthcare organizations still monitor their revenue cycle through periodic reports and manual reviews—often waiting until end-of-month summaries or deep-dive analyses to understand issues like claim denials or revenue leakage. This reactive approach means that by the time problems are detected, they may have already compounded, leaving teams scrambling to resolve them.

This is compounded by the unstructured data problem. There are tons of documents, faxes, claims, and even internal conversations that could help unlock operational insights that are difficult to incorporate into any data project (which are already notoriously risky).

Why AI Can Help

• Real-Time Monitoring: AI systems continuously monitor both structured and unstructured data, flagging anomalies as soon as they occur and enabling teams to address issues before they escalate.

• Unlocking Hidden Insights from Unstructured Data: AI-powered language processing transforms messy, low-ROI unstructured data into actionable insights by revealing hidden patterns and extremely specific insights for your organization.

• ** Proactive Interventions:** An AI model might not always provide perfect answers, but it can contextualize your data to suggest reasonable action steps. This helps kickstart team responses and frees up valuable time for forward-looking initiatives—an outcome every organization desires and finds much more achievable with AI.

• Enhanced Root Cause Analysis: Combining structured metrics with the "narrative" of your data, AI can help teams understand the inputs of problem areas much more quickly. Again, this is an accelerant to your team doingwhat it does best.

Conclusion

I understand that we've worried about AI completely taking over tasks in healthcare. Technology companies/commentators have done a poor job of being on friendly terms with real healthcare on the ground, such thatstaff and leaders can trust it.

But most folks are feeling that they can no longer resist this trend. I think that's right, but I hope that I've given you some mental models of how AI can be used in areas where it excels, as well as helping staff rather than scaring them.

I have hopes that we can adjust course and get healthcare on board in the right ways that help patients. Sure there are much deeper problems in our industry, but I think the first step is learning how to survive in a sustainable way.

And fortunately, that path is more clear now than it has been for a long time.

About Me

Hi, I'm Devon Mobley. I'm the Chief Growth Officer at Calvient, a Practical AI platform for healthcare.

I have worked in specialty care designing and building workflows with software as well as in Big Tech as a senior engineer.

My mission is to serve the real people doing the real work in healthcare. For me, innovation happens at the margins in healthcare, so I aim to bring people together to push the boundaries: slowly, practically, but progressively.

If you would ever like to chat about AI, strategy, or healthcare technology/operations, I am an open book!