Think AI Podcast - When Data Meets Health: AI in Healthcare That Actually Works | Ep. 2

Episode Date: March 11, 2026

The U.S. healthcare system loses $260 billion a year. Not from malpractice. From paperwork, billing errors, and missing data.In this episode, Dave Goyal — founder of Think AI (top 1% of 500,000 Microsoft partners), Harvard-trained in AI for healthcare — shares 15+ years of real healthcare projects. No theory. No hype.You’ll hear how he built a claims system from zero that generated $100M+ in annual revenue. How he designed data strategies for entire health networks. How he helped non-profit hospitals turn buried patient records into life-saving insights.Plus: six back-office AI use cases any healthcare org can start today, a straight-talk segment on privacy and HIPAA, and an AI Tip of the Day you can try in two minutes.Next up: Episode 3 — AI in Manufacturinglinkedin.com/in/davegoyal | davegoyal.com #HealthcareAI #RevenueCycleManagement #DataStrategy #ThinkAI #AIpodcast

A hospital loses money every single day, not because of bad nurses, because of bad data. A claim gets denied. Nobody knows why. A patient record is missing one field. The bill sits for 60 days. A surgeon performs a procedure. The code is wrong. Revenue disappears. The U.S. healthcare system loses over $260 million a year.

every year just from revenue cycle problems not malpractice not lawsuits paperworks billing errors missing data two hundred and sixty million dollars welcome to the think i podcast each week we talk about the most exciting AI research tools case studies and more i'm your host Dave Goir and I've been working behind the scene in data and AI for over 30 years, whether you are an AI expert, skeptic or something in between, this podcast is for you. I have spent over 15 years inside healthcare organizations, not as a doctor, not as a nurse, as the guy they call when the data is broken. I have helped build claim systems from zero. I've designed data strategies for the entire health network.

I have worked with non-profit hospitals drowning in patient records. They couldn't make sense of it. And I have seen up close what happens when healthcare organizations doesn't get data right. And when they get it right. People get paid faster. Patients get better care. And the back office, being the cost center, it changes. from that into a competitive advantage.

And that's what this episode is all about. Welcome back to Think AI podcast. I'm Dave Goyer. And today we are going deep into healthcare and AI. Real stories, real projects, and real numbers. If you run a healthcare organization or work in one, this episode will change how you think about your back office. If you are not in healthcare, stay with me anyways.

Because the lessons here apply to every business city, on data it doesn't know how to use. Let me get into it. Let me take you back to where it all started for me in healthcare. I was working as a consultant. This was early in my life in the US. I had my green card. I was hungry and I wanted to prove myself.

I got an engagement with a large medical device organizations, one of the biggest in the world, global company, thousands of employees making devices that go inside the huge. human body, catheters, surgical instruments, things that save lives every day. My role? I started as a web development consultant, basic stuff, building internal websites, managing environment, but here's what I learned early in my career. If you show up and solve problems, not just the ones they hired you for, but the ones nobody else wants to touch, doors open. So I started solving problems,

nobody is asking me to solve. I noticed that teams were siloed. Engineers didn't talk to IT. IT didn't talk to operations. Nobody cared and nobody shared data when anyone. So I started connecting the dots. I moved from managing websites to managing environments, from environments to projects,

from projects to overall programs. Within the short amount of time, I was operating like a program director, handling multiple projects and programs across teams, learning about infrastructure, learning about intranet, and this is the big one, learning about data analytics. This is where my data analytics foundation was formed up. See, inside this medical device company, there was data everywhere,

production data, quality data, supply chain data, sales data, and nobody was connecting it. Each department has its own spreadsheet, their own database, their own version of truth sounds familiar. So if you work in healthcare or really any industry, you are probably nodding right now. Because this is the number one problem I see in every organization. Not a lack of data, but the lack of connected data or connected system in this case.

Let me put it this way. Imagine you have a puzzle. Thousands of pieces. But each department has 100 pieces and nobody is sitting at the same. table putting it together. That's what most healthcare organizations look like from a data perspective. Every system has a piece on the picture. The EHR has clinical data, the billing has financial data, the scheduling system has operational data, the lab has diagnostic data but nobody is seeing the full picture, nobody. And that's where I come in.

That's what I've spent my career doing putting this puzzle together. Now that's the The first medical device engagement taught me something critical. It taught me that healthcare is not like any other industry. In manufacturing, if your data is wrong, you might make a bad product. Now that is bad and you might lose money. But in healthcare, if your data is wrong, someone might get the wrong treatment, might even die. Someone might not get treated at all.

The stakes are completely different and risky. The data has to be right, not mostly right, right. And that standard, that relentless pursuit of accuracy became the foundation of everything. I built in healthcare from that point forward. Alright, let me tell you about the project that changed my career. After that first engagement, I went deeper into healthcare. I kept learning, I kept growing. Eventually I started working with another medical device organization.

company was having a unique position they were building a brand new claim business from scratch not fixing an old system not upgrading softwares but building a claim operation from the ground up think about what that means when a patient gets in and get a medical device let's say a knee brace or a surgical support someone has to build an insurance company that's a claim the claim has to have the patient information, the right procedure code, the right diagnostic code, the right pricing, the right provider information. Any of this is wrong, the claim gets denied. The company doesn't get paid. Now multiply that by hundreds and thousands of claims per month.

That's what we are building for. We use the full Microsoft technology set like databases, reporting tools, integration layers, custom applications. But here's what we are building. But here's what this project, which is different from every other tech project. We didn't start with technology. We started with the process. I sat down with the business teams. I asked them, walk me through what happens from the moment a device is shipped to the moment you get paid. And what they described was a mess.

Manual spreadsheet. Phone calls to insurance companies. Paper faxes. Yes. Paper faxes. This was healthcare. Paper faxes. or the faxes were everywhere.

There were delays at every step. Claims would sit in for weeks. Denial would pile up. Nobody could track where the bottleneck was. So we mapped the entire process end to end, from device shipment to cash in the bank. Then we build the system around that process. Three things we got it right. Number one, we automated the intake.

The intake process was when the claim comes in, and the system very well, validated every field that patient info, codes provided data or something was missing. It flagged immediately, not after 90 days immediately. Number two, we built real-time dashboards. So I did write a book on real-time business intelligence mastery. It is for healthcare and manufacturing leaders who can use this. We build the real-time dashboards. The leadership could see at any moment how many claims were in place.

process. How many were approved and how many were denied and why? Now the why part. That's where the magic is because you know why claims are getting denied. You can fix the root cause, not just resubmit the claim. Fix it. Fix the problem with the source. So we build a denial system workflow. Every denied claim went into a queue. It was prioritized by some dollar amount and likelihood of that recovery and the team knew exactly what to work on first. The result? That company went from zero revenue to $100 million a year, processing 100,000 claims per month, from nothing to nine figures using data and technology. But here's what I want you to take away from this story. It's not the technology. It's not the platform, but it's the

approach. We started with the business problem. We mapped the entire process. We identified the bottlenecks and then, only then, we build the technology to fix them. Most companies do it backwards. They buy the software first and then they try to fit their process into it. That's why most project fails. We didn't fail and that's why the success rate across all our projects sits in at about 95%. We are pretty proud of it. Now, that claim project taught me something about healthcare that shaped everything else I did. Healthcare doesn't have a technology problem.

It has a data strategy problem. And that leads me to my next story. So after the claim success, I started getting calls from larger healthcare organizations, not just medical devices company, but the full health systems. Hospitals, clinics, multi-site networks, and the question always remain the same.

Dave, we have this data, we don't know what to do with it, every single time. Same question, different organization. Now let me tell you what all this data looks like inside a typical healthcare organization. Let's say you have an EHR, that's an electronic health record system. Think of it as the digital version of your medical chart. It has your diagnosis, your medications, your lab results, your visit history. Then you have billing system that tracks claims, payments, denials and adjustments. And then you have scheduling system that tracks appointments, no shows, cancellations and wait times.

Then you have HR and payroll because you need to know how much it costs to run the staff. And then you have supply chain because every bandage, every syringe, every device costs money. And then you have the quality matrix, patient satisfaction scores, readmission rates, infection rates. Now, as you can see, these are like six or seven different systems. This is like a value stream of an organization. In most healthcare organizations, it's more like 15 to 20 different systems. And here's the problem. None of them talk to each other.

team cannot see the clinical data, the clinical team cannot see the financial data, the executive gets a report once a month and everything is outdated the moment it prints or sent it through an email in this example. So when a healthcare organization calls me and says, help me figure this out, that's what we are solving, not a software problem, it's really a strategy problem. is what building a data strategy actually looks like step by step step one you assess you audit you sit with every department you ask what data you have where does it live who uses it every decision that you make with it and this is the key question what

decisions can you not make because you don't have the right data data could be good bad, wrong or right. We will deep dive into it at some point. The last question is where the gold is. Step 2. You map the data landscape. So we started with an assessment and audit. Now we are mapping the data landscape.

Every system, every database, every spreadsheet, every report, your document where the data is created, where it moves, You document everything and you also document where it breaks. And trust me, it breaks a lot. Patient names are spelled differently. Their phone numbers, addresses, demographic changes in different systems because of different point in time the data is entered. Dates are formatted differently.

Could be because of geolocations and how the person is handling it. Coats doesn't match. Records are duplicated. We call this data. quality and in healthcare poor data quality doesn't just cost money it could put a patient in danger at risk step three so we saw audit and assessment we looked into mapping the the landscape third is you define the architecture this is where you decide how are you going to bring all the data together what tools you are going to use what is the single

source of truth going to look like. We typically build what's called a data warehouse or a data lake house in the modern world. Think of it as one big clean, organized library where all of the data from all the system lives in one place, structured, govern and secure. Step four, you build the analytics. This is where the value hits. Now that you have clean connected data, You can start asking questions you could never done it before. What procedures are most profitable? Which ones are losing money? Which providers have the highest patient satisfaction and why?

Which payers denies the most claims? What's the pattern? How many patients are at risk of free admission within 30 days? Can we intervene early? Where are the bottlenecks in scheduling? Why do patients wait 40 to 40? minutes when they should only wait for 10. We have been sitting on this data for years. We had no idea how it could tell us this much. That's the power of a data strategy. It

turns raw information into answers and then answers into action. Now let me take you to a different healthcare organization. Non-profit hospitals. Non-profits are a different world. They don't operate for profit, as the word says. Their mission is to serve the community, often that means serving the most vulnerable patients, patients who are uninsured, patients on Medicaid, patients with complex conditions and limited resources. The challenge? Nonprofits have the same data problems as for profit organizations, but they have far fewer resources to fix them. Smaller IT teams, tighter budgets, older systems. I worked with two non-profit healthcare organizations,

Both had the same core problem. They had an EHR, they had the patient data, mountains and mountains of it, years and years of patient records, but they could not use it or use it intelligently. They couldn't answer the basic questions. How many diabetic patients do we serve? How many are overdue for the screenings? Which patients have been to the ER more than three times? And why? These are life and death questions. And the data to answer them exist like it was sitting right there in the EHR, but it was locked, buried, inaccessible.

So here's what we build. We started by connecting the EHR data to a modern analytics platform. We build what I call it a patient intelligence layer. Think of it as a lens. You put it on top of your existing data and suddenly you can see things you couldn't see. see before. Let's talk about a few use cases. Use case 1. Chronic disease management. We built a dashboard that showed every patient with a chronic condition, diabetes, hypertension, heart disease, COPD. For each patient, the dashboard showed, when was their last visit?

Are they current on medications? Have they completed all their screenings? Are they having a a risk for complications. The care team could look at that dashboard every morning. They could see right there, which patient needed outreach, which ones were falling through the cracks. Before this, they were guessing, pulling charts manually, making phone calls based on memory or maybe some rough notes somewhere. After this, it was precise. It was proactive, patient got calls before they get worse. That's what data does. It turns reactive care into proactive care. Use case 2.

Emergency Department Utilization. One of the organization had a major issue. The small number of patients were using emergency department over and over again. Sometimes 10, 15, 20 visits a year. Now, that's not good for the patient. The ER is not the right place for ongoing care. It's expensive, it's stressful. It treats symptoms, not the real.

root cause. And it's not good for the organization's every unnecessary visit, especially ER visit, costs thousands of dollars. So we analyzed the data. We identified the top frequent users, we looked at their diagnosis, we looked at their demographics, their social characteristics, their access barriers. What we found was not surprising. It was powerful. Many of these patients didn't have the primary care physician. Or they had one, but couldn't get an appointment. Or they had transportation issues or they were dealing with mental health challenges

along with the physical ones. The ER wasn't failing. The system around these patients was failing. So the organization used our data to create outreach programs. They connected these patients with care coordinators. They arranged transportation. They schedule primary care visits

and ER visits dropped. Patient outcomes have improved and the organization save real money, not because of Fancy AI, because of connected data systems used for this purpose. Use case 3, quality reporting and compliance. Non-profit hospitals have to report quality matrix to the government, CMS,

which is center of Medicare and Medicaid services, which requires it. Things like what's your readmission rate, what's your patient satisfaction score, and what's your infection rate after surgery. These reports used to take weeks to compile. Staff would pool data from many systems, copy into spreadsheets, cross-reference manually, check for errors, it's painful, let me tell you. It was slow and it was completely into errors or error prone, as you would say.

We automated the entire process. The data flowed from the source systems into the analytics platform. From there, it went into reports which were generated automatically. And the compliance team could review and submit it in a fraction of time. What used to take three weeks, now it took two to three days and even faster in some cases. What used to require four people required just one. That's not magic. That's the data infrastructure or data applications.

done right. All right, let me switch gears. So far I have told you stories about what we did with data and analytics, connected systems, building dashboards, cleaning up the messy information. Now I want to talk about where AI takes all of that into the next level. And I want to tell you why I'm qualified to talk about this. After years of doing this work in the field, building claims systems, designing data strategies, working inside hospitals, I realized something. The technology was changing way faster than my experience alone, which could keep up with it. So I did what I always do.

I went in and learned. So I studied AI in healthcare at Harvard, not because I needed a certificate on my wall, but I needed to understand at the deepest level where AI was heading in the industry. What was real? What was hype and where it will change? the patient care forever. That experience connected the dots for me between the data work that I have been doing for years and the AI capability that were now becoming possible.

Everything I'm about to share with you comes from that combination. 15 plus year in the trenches and a deep study of where this technology is going. And I want you to be really clear about something. AI in healthcare is not about replacing doctors. It's not about replacing nurses. It's not even about doing robotics performing surgery or robots performing surgery. That stuff gets the headline. Biggest value of AI in healthcare is in the back office.

Let me walk you through the use cases. These are real. These are happening and any healthcare organization can start implementing them. So make a note of it. Connect with me if you need more information. on this. Use case one. AI powered claims management. Remember the claim system I build? Imagine that system, but smarter. Today, AI can take a look at claims before it's submitted. It can compare it again thousands of payer rules. It can predict before you even hit send. Whether that claim will

be approved or denied, think about that. You know the answer before you ask the question. That's Wonderful. If the AI can predict a denial, it will tell you why. Wrong code, missing authorization, incomplete documentation, and it gives you the fix right there in real time. Healthcare organizations that use AI for claim predictions are seeing denial rates drop by like about 20 to 30 percent. That millions of dollars recovered or saved every year. Use case 2, automated medical coding. Every procedure a doctor performs gets a code. Every diagnostic gets a code.

There are thousands of these codes. A human coder reads the clinical documentation. They interpret what happened. They assign the code. It's not tedious. It's not complex. The coder makes mistakes. Not because they are bad at their jobs,

but because there are too many codes, too many rules. and not enough hours in the day. AI can now read clinical notes. It understands the language. It assigns the right codes. It does it in seconds, not hours. And here's the key. AI does not replace that job of the coder.

It does the first pass. The coder reviews and approves. Human in the loop. AI does the heavy lifting. Humans make the final call. One health system reporting, I would say about half a million dollar in coding costs just from the AI coding or AI doing that first pass. Use case three, prior authorization automation.

This is the one that drives everyone crazy. Before a patient can get certain treatment, the insurance companies has to approve. That's called prior authorization. The process is slow. I've been there. It requires phone calls, faxes. Yes, still faxes, forms, follow-ups. And while everyone waits, the patient waits too,

sometimes dates, sometimes weeks, and even more. AI can now automate large portions of this process. It gathers the clinical documentation, it fills out the forms, submit it to the pair electronically, and tracks the status. What used to take a nurse 40, 45 minutes per authorization, like five minutes, multiplied that by hundreds of authorizations per day. That's your nursing staff back. Spending time with patients instead of fighting with insurance companies.

Don't you want to have it? Use case 4. Predicting patient no shows. This one is really simple but powerful. Patient doesn't show up their appointment or they do not show when their appointment is about to happen. that slot is wasted, the provider loses revenue, another patient who needed that slot didn't get it. AI can analyze that patterns. How?

Which patients are most likely to miss their appointment, based on history, based on demographics, based on day of the week, based on how far they live from the clinic? Once you know, who might show up, you can act. Send a reminder, offer a telehealth option, double book them. One organization reduced no shows by 25% just by smart scheduling which is powered by data and AI. Use case 5. Readmission risk prediction. When a patient leaves the hospital, there is always a risk that they will come back within 30 days.

That's called readmission. It's bad for the patient. And under certain government programs, hospital gets penalized financially for high readmission rates. AI can look at the patient record at discharge. It scores the risk, high risk, medium risk, low risk. If the patient has high risk, the care team knows. They can schedule a follow-up call, arrange home health services, make sure medications are filled. It's not about predicting the future.

It's about acting on what the data is telling you right now. Use case 6. Intelligent Document Processing Healthcare runs on documents, referral letters, insurance cards, consent forms, lab results from outside providers, and most of this arrive as scanned images, PDFs, faxes and someone has to read them. Manually enter the data into the system. AI with optical character recognition and natural language programming, the two short forms should be OCR and NLP, can now read these documents.

Extract the key information, populate the right field in the EHR automatically. Staff spend less time typing, more time caring. So six use cases, all back office, none of them involved replacing a single clinician. Don't you see that? It's not taking jobs. Every one of them saves money, saves time, reduce errors, improve overall patient experience. And of course, the back office will be happy too. This is what AI in healthcare actually looks like.

Not science fiction. Practical, miserable and real. Now I know what some of you are thinking. Dave, this all sounds great. But what about privacy? Great question. I'm glad you're asking it. Because if you're not asking it, you should be.

Health care data is the most sensitive data that exists, your medical records, your diagnosis, your medications, your mental health history. In the US, we have a law called HIPAA, the Health Insurance Portability and Accountability Act. It sets strict rules about how patients, data can be stored, access, shared and used. AI does not get a free pass on HIPAA.

Every AI system we build in healthcare must be HIPAA compliant, period. That means data is encrypted, access is controlled, audit trails are in place, patient consent is respected. And now here's what I'm going to tell the skeptics. The risk is not AI. The risk is bad implementation. of AI. So if you deploy AI without governance, without rules, about what data it can access, how it is trained, who reviews it, decisions, then yes, we have a problem, you have a problem.

But if you deploy AI with proper data governance, with human oversights, clear policies, you actually make patient data safer, not less safe, but safer. Because right now in most Health Care Organizations, data is scattered across 20 systems, or even more in some cases. People email spreadsheets with patient information, which is in text files, as you can see. Paper records sit in unlocked rooms. That's the real privacy risk, not AI. The current manual, fragmented, uncontrolled mess is the problem. AI, done right, centralizes data, control access.

logs every interaction, alerts you when something goes wrong. So if you're a skeptic, I respect that. Stay like that. Ask hard questions, demand compliance. But don't let the fear of AI stop you from fixing the real problems because the real risk is doing nothing. All right, time for AI tip of the day. Last episode, I showed you how to build your first AI assistant in 10 minutes.

I hope you tried it. If not, go back and try it. It works. Today's tip is specifically for anyone who works in healthcare, but honestly, it applies to any business that deals with a lot of documentation. Today's tip, use AI to summarize long documents instantly. Here's the scenario. You receive a 50-page policy document from an insurance company

or a 100-page compliance report or a dense research study. Nobody has time to read all of that, but you need to know what's in it. Step 1. Open any AI tool like JetGPT, Claude, Copilot, Gemini, your pick. Step 2, upload the document or copy-paste the text. Step 3, type this prompt. Summarize this document in plain English.

give me five most important points. For each point, tell me what action I should take. Keep your language simple. No jargon. That's it. In 30 seconds, you have a clear summary with action items. Now, here's the advanced version. After you get the summary, ask a follow-up question.

What are the three biggest risks in this document that I should, be aware of. Or, how does this policy compare to the previous version we had? Or, write me one paragraph email to my team explaining the key changes. You just turned a 50-page document into an email in under two minutes. For the AI curious, this is the simplest way to see how AI save you real time. Try it once. You'll be hooked. For the enthusiast, start building this into your daily workflow. Every document that comes across your desk, run it through AI first. For the skeptic, read the original document after you read the summary.

See if the AI got it right. I think you'll be surprised. That's your tip. Two minutes, one document. Try it today. Now here's what coming next. In episode 3, I'm taking you inside the world I love, manufacturing. Just like healthcare, manufacturing is sitting on a gold mine of data.

It doesn't know how to use, plain and simple. I'll tell you about the medical device companies we worked with where a single production error could affect thousands of patients. I'll talk about how we helped organizations use data to predict equipment failures before they happen, how we build real-time. production dashboards that cost waste and increased output. I will share what happens when you connect your shop floor data to your front office, when the machine starts talking to the balance sheet.

Manufacturing and AI is not a nice to have anymore. It is survival. Most manufacturers are still running on spreadsheets and gut feeling. I'll tell you how to change that step by step. That's episode 3. Don't miss it. If this episode helped you, if something I said made you think differently about your data, your back office, or what AI could do for your organizations, here's what I would love you to do.

Subscribe. Share this with someone in healthcare who needs to hear it. A CIO, a CFO, a revenue cycle director, a practice manager, and drop me a message. Tell me what challenges you are facing with data. I read every single one. You can find me on LinkedIn, Dave Goyle or at Dave Goyal.com. I'm Dave Goyer and this is the Thinkye podcast. I'll see you in the next one.

You have been listening to ThinkYi podcast with Dave. Take one idea from this episode and turn it into action.