Years ago, I was gifted a laser measuring tool. Want to know the distance from one wall to another? Simply hold this tool at one wall, press the button, and voilà – measurement. At the time, this was new and innovative. Measuring dimensions of a room was now fast and painless compared to the old method of using a floppy tape measure and moving furniture. While I don’t need to use it often, it is a valuable tool to have for specific circumstances. However, I never use my fancy new tool to drive a nail, turn a screw, or cut a piece of wood. I have other tools for those needs; tools that have been around for ages, have stood the test of time, and are still the best tools for their specific jobs.

The shiniest new tool in healthcare revenue cycle today is Artificial Intelligence (AI). I recently attended a conference where the majority of vendor booths touted the use of AI in their solutions. When looking deeper into their offerings, many are utilizing automation in the form of RPA (robotic process automation, or just “bots”) and labeling it AI. During a presentation by a long-time industry professional, the terms RPA, AI, and Machine Language (ML) were used interchangeably. At one point the words, “or whatever term you want to use,” were spoken, implying that RPA and AI/ML are the same thing. They are not. While RPA tools sometimes make use of AI, they in and of themselves are not AI anymore than my laser measuring tool is a hammer.

Confused? Let’s start with a quiz. Of the technology applications listed below, identify which are AI:

1. Analyze remittance data to identify denials trends.
2. Dynamically monitor data and notify key personnel when a known problem scenario presents itself.
3. Automatically send open claim information to a claim status vendor and then import the results.
4. Utilize a bot to login into a payer portal and check a patient’s eligibility.
5. Create an appeal letter with the click of a button.

If you answered that none of these are AI, you may stop reading now as you have a clear understanding of what is and is not AI. If, however, you are surprised that none of these are AI, allow me to explain. #1 is simply data analysis. Powerful analytics tools have existed for decades and are still the right solution for analyzing past and current data. #2 is an automated monitoring tool. A programmer can create a query, schedule it to run, and then email folks when necessary. Third party monitoring tools make this an easy task. #3 is an integration. This can be performed typically via a batch/electronic data interchange (EDI) process or via the vendor’s application programming interface (API). This technology has existed for decades with REST APIs being the preferred method of integrating systems for the past 10 years. #4 is RPA. As long as the payer’s portal permits the use of bots, this is an acceptable use of RPA. However, eligibility can typically be checked using more conventional methods (such as #3) that tend to be far more stable than the ongoing babysitting that bots require. #5 is simple programming. It is not automation, and the furthest thing from AI on the list.

So, what is AI? Put simply, AI is prediction. The AI powering the suggestions presented by Netflix and Spotify is attempting to predict what you want to watch and listen to next. The AI in self-driving vehicles is taking in enormous amounts of data and attempting to predict what a stop sign looks like. The most talked-about AI right now, ChatGPT, is simply attempting to predict what the next best word is in a sequence. It has the benefit of having been trained on an enormous amount of data, which gives it the appearance of being “intelligent”, but at its core it is really just using complex math to make predictions.

When looking through the lens of AI as prediction, it becomes obvious that many other technologies are not AI. A technology that estimates when a payer will pay a specific type of claim is likely using AI. It has been trained on millions of historical claims and can now predict outcomes with an acceptable accuracy rate. A technology that uses a bot to automate the rebilling of a claim is not predicting an outcome, therefore it is not AI.

So why is there so much confusion around AI and RPA? If AI is for prediction, why is the term being applied to automation? Do some people simply think the A in AI stands for Automation? Most likely the confusion is due to a lack of understanding of what AI truly is, combined with a strong desire to use the term for marketing purposes. We saw the same scenario play out with the advent of cloud computing. Amazon Web Services (AWS) was the pioneer, at least publicly, of cloud computing, just as ChatGPT has become synonymous with AI. Technology vendors were eager to advertise their solutions as being “in the cloud” or “cloud-ready”. Many at the time didn’t even really know what that meant, and they assumed because they had a website, they were a “cloud solution” when most were not. Today, most technology solutions are cloud-based, but you’ll never hear anyone brag about it. It’s simply a foundational element of most technology services. Soon, the same will be true of AI.

Here is a final quiz to test your knowledge. Which of the following are AI:

1. Utilize past COB denials data to predict if a claim will be denied due to a coordination of benefits issue.
2. Using past clean-claim payment data and patient demographics, predict how many days it will take a specific payer to pay a radiology claim.
3. Predict how often a promise-to-pay remit or claim status check actually results in a payment.

Hopefully the word “predict” in all three examples made this an easy A for you. All three examples use past data to predict future outcomes. That is AI. The results from those predictions could be used in analytics, and yes, even instruct bots to perform a task.

Photo: zhuweiyi49, Getty Images