What is a "System of Intelligence"?

Posted by Alex Smith on Dec 14, 2018 12:31:28 PM

 system of intelligence:  noun      sys•tem of in•tel•li•gence

1. an AI solution that works across datasets and systems, growing faster and smarter over time 

2. the technology by which the Digitize.AI team lives out their values and fulfills their company mission 

"Ok, I get the Webster dictionary definition... but what is a system of intelligence, really?"

To classify as a system of intelligence, the machine (or platform) needs to fulfill a certain set of attributes. These attributes include the capability to take in data from multiple sources, process and analyze that data, and communicate desired outputs back into multiple systems if needed . These systems use artificial intelligence and machine learning models to process the data, so they are continually growing smarter over time.

More simply, a system of intelligence does a specific job, end to end, by working within the systems and workflows that you currently have. It's the inverse of traditional software in that it does not introduce new interfaces or interaction points—in fact, it works to remove them entirely.

Image source:  Jerry Chen's, "The New Moats"

"OK. So, what's an example of a system of intelligence at work?"

A prime example is Lia™, our system of intelligence for faster and smarter prior authorizations in healthcare. To understand Lia's capabilities and value, let's take a look at "life before Lia" and "life after Lia."

Life before Lia
First, some context—"prior authorization" refers the process by which a hospital pre-confirms that an insurance company will reimburse the hospital for a specific treatment, for a specific patient. If your doctor were to schedule you for an MRI or heart surgery, the hospital would send your financial and clinical information to your insurance company to double-check that you're covered for that given treatment (and to create a paper trail to be absolutely sure the insurance company will write a check for the cost when the care has been delivered).

Historically, the prior authorization process has been reliant on manual processes and lots of human labor. Hospitals employ back-office teams that gather data from the hospitals electronic health record (EHR) and then submit that information to insurance companies via phone, fax, and web portal with the goal of securing an "authorization."

This process is imperfect, of course. The fact that the key systems systems of record (e.g. EHRs, insurance systems, etc) and systems of engagement (web portals, phone, fax) are unintegrated means that the prior authorizations rely on manual work and data-sharing. This inherently causes delays, error-rich submissions, and imperfect data-sharing can delay treatments, create headaches for patients and clinicians, and cost billions of dollars each year in administrative inefficiency.

Life after Lia

Lia integrates via APIs to (1) hospital EHRs (Epic, Cerner, etc) and (2) Payer's back-end systems to seamlessly, accurately, and immediately submit prior authorization requests to payers at the moment they are needed, and to later return the response from the Payer directly into the EHR system. Think of Lia like a behind-the-scenes data bridge, automatically moving data between Providers and Payers and employing machine learning to know what data are needed—and when—in order to reduce friction in the process. (To see Lia in action, click here for a demo video). Lia cuts administrative costs, eliminates errors, and executes a specific job (e.g. prior authorizations) by orchestrating workflows and knowledge-sharing between and across existing systems of record and systems of engagement.

"Got it. So then how do the other buzzwords I've heard about—AI, ML, RPA, APIs—fit in here?

They are all potential working components of a system of intelligence. If you think of a system of intelligence like a human body, these other things make up the brains and legs.

Legs: Technologies that can make processes run quickly. 

  1. Robotic Process Automation (RPA) – the process of creating software robots to perform tasks. Think of RPA as your solution for manual task completion, like entering in data on an EHR screen or clicking through a website to submit a request. It is important to note that RPA technology is not artificial intelligence technology. These tools are merely for operating manual tasks, but can by helpful in allowing AI platforms to execute on learning. 
  2. Application Program Interface (API) - these are effectively pipelines or superhighways that run between systems, behind-the-screens, to quickly share data. They're bi-directional i.e. they can gather and publish data between systems, but are only as smart as the logic that tells them what data to share, and when.
Brains: Technologies that can make those quick processes run in the right direction, at the right time.

  1. Machine Learning (ML) - is the process of teaching the software to learn over-time. This ability to learn comes from data that is fed into the system. Think of machine learning as the brain behind the operation. Machine learning consists of models, that over time, learn from previously analyzed data and constantly improve decision making and analytic capabilities. An example would be the system learning to predict the likelihood a prior authorization request will be approved based on data such as historical success rate, patient demographics, and timing of the request.
  2. Artificial Intelligence (AI) - is the theory and practice of developing computer systems to perform tasks that normally require human intelligence.

("Wait... aren't machine learning and artificial intelligence the same thing?")

The short answer, no. The longer answer - machine learning is part of artificial intelligence. All machine learning is AI. But not all AI has machine learning. Machine learning elevates an AI model to allow the system to make decisions in real-time based on data inputs. Machine learning and automation combined qualify a platform as a system of intelligence capable of navigating from record systems and engagement systems.

To put into context for health insurance payers, machine learning allows a system to make decisions in real-time whether to automatically approve a prior authorization request from a provider based on historical data or to escalate the request to a care management representative for additional review. Machine learning as part of the holistic AI technology empowers the system of intelligence to make decisions based on historical data inputs and outcomes.

Closing the loop on this "brains" and "legs" analogy: If you only have "legs", you're doing a lot of things quickly... but not necessarily intelligently. If you only have "brains", you have all the insight but no vehicle for enabling it to solve a problem. You need both. And a system of intelligence marries them together.


Why are systems of intelligence valuable?

A system of intelligence is a platform designed to provide a holistic approach to solving a problem. This type of system uses artificial intelligence, infused with machine learning and combined with data, to continually learn and grow more effective over time. Systems of this nature combine multiple technologies to predict outcomes and act accordingly. These types of predictions allow a system of intelligence to reduce the administrative workload often associated with revenue cycle management and care management by submitting requests, making decisions based on historical data, and engaging with the record systems and engagement systems necessary to complete the process.

Systems of intelligence are not simple automation designed to perform manual tasks. Instead, the machine learning embedded in these systems provides a level of decision making traditionally reserved for humans. Benefits of a system of intelligence include a decrease in human labor, maximized productivity, and decreased costs.


Lia™ and Mia™ as Systems of Intelligence

Again, we've created our technologies from the ground-up with this system of intelligence approach at the core.

Lia™, our System of Intelligence for the prior authorization process and Mia, our solution for utilization management, use AI infused with machine learning. Lia™ and Mia™ both tackle repetitive, mundane tasks intelligently, growing smarter over time. 

Lia™ is able to submit prior authorizations requests behind the scenes, extract necessary information from the EHR system, and then submit to the appropriate payers. With each submission Lia™, using machine learning, learns specific payer approval rates and grows smarter for more successful future decisions. 

Our other System of Intelligence, Mia™, uses the same machine learning technology to prioritize the prior authorization submissions on the care management side of the process. Mia™ will then instantly approve cases that meet payer rule criteria or escalate appropriate cases for further review by an employee. 

Lia™ and Mia™ both exhibit qualities of AI by performing tasks that currently require human intelligence and continue to learn through their machine learning modules, growing more effective and appreciating in value. 


Alex Smith

Written by Alex Smith

Topics: intelligent automation, artificial intelligence, pre-authorization, revenue cycle management, Healthcare IT, Healthcare, pre-certification, revenue cycle, "AI for Health Plans", system of intelligence