The Advantages of AI for Claims
Claims is integrating AI in several functions to help boost efficiency and substantially reduce costs.
Confidential File Summaries
The team is working with an outside vendor, CLARA Analytics, which will use its intelligence platform to help automate the creation of confidential file summaries, documents that present key information on a claim:
- Treatment facts
- Policy limits information
- Deductible information
- Dates of loss
- Names of involved parties and treaters
- Opinions of insureds
- Experts and defense counsel
- Analysis of applicable venue, judge, and plaintiff counsel
These summaries can take Claims Specialists, at times, up to 8-10 hours to create. CLARA can help reduce that time to minutes.
Automating confidential file summaries frees up time for Claims Specialists to work on medical and legal issues on files. The process also allows Claims staff to spend more time interacting with our insureds and business partners, which helps ProAssurance achieve our Carrier of Choice goals as well as our customer service goals.
Medical Record Summaries, Deposition Summaries, and Medical Chronologies
Claims also partners with several vendors that use AI to help facilitate medical record and deposition summaries and medical record chronologies.
Depending on the number of records or the length of a deposition transcript, it can take attorneys, paralegals, or Claims staff 10 or more hours to complete these summaries. Law firms also charge $110-$250 an hour for this service.
By using AI the work is performed in minutes, leading to a substantial reduction in costs and greater production in less time.
Claims Predictive Modeling
Working with the company Growth Protocol, Claims is additionally developing a claims predictive modeling product, which will use AI to collect ProAssurance data and court data across the country. That data will help predict specific factors related to cases, such as average settlement values, average range of verdicts, and average duration of cases. The product will also gather insight into courts’ tendencies and plaintiff counsel playbooks (what motions they file, cases won, cases lost, experts used).
Having such information quickly available to the defense team will be extremely valuable in helping develop defense strategies on files, including valuation of cases.
An Update on Phyllis, Eastern’s AI-Powered Associate: Two Years of Success
It’s been two years since Eastern Alliance introduced Phyllis, an AI-powered digital team member, to its mailroom operations. Since her debut on April 26, 2023, Phyllis has transformed how Eastern handles insurance documents and claims processing. What began as an innovative solution to automate routine tasks has evolved into a sophisticated system that now handles almost all of Eastern’s Claims mail either fully automated or with minimal human intervention.
Background
Phyllis emerged from a partnership with Roots Automation, creators of the world’s first language-oriented platform enabling insurance-knowledge workers to create AI-powered “digital coworkers” using speech and text. Eastern began exploring the platform in September 2022, and after an extensive training period led primarily by Jennifer Zimmerman, Eastern’s Business Systems Manager, Phyllis was ready to transform the mailroom’s operations by April 2023.
Over the course of several months Roots Automation worked with Zimmerman as she explained exactly how the mailroom associates did their job, step by step, indexing thousands upon thousands of emails—primarily vendor bills, correspondence, state forms, and medical reports. She narrated how all of them needed to be reviewed and indexed to the right document type and sent to the right person.
As Zimmerman painstakingly explained the work, Phyllis began to learn.
She came programmed with key capabilities needed to effectively manage the documents and systems commonly found in an insurance mailroom. She holds an extensive amount of insurance-specific data and process knowledge through Roots Automation’s proprietary InsurGPT AI model—a large language model (LLM) created specifically for insurance.
She is built with AI that is capable of reading screens, documents, forms, and data like a human. She can dynamically identify and interact with objects on a computer screen, such as buttons, icons, form fields, and field names.
Leveraging advanced Natural Language Processing (NLP) to collaborate and communicate with people around document exceptions, Phyllis is continually adapting and learning.
She captures transactional and meta-data through interactions with people. If she is not “confident” in her ability to process an item, she can send that item for “review,” where one of our human team members can handle it for her. Phyllis undergoes monthly performance retraining, with the goal of increased confidence and ongoing improvement—meaning she gets smarter every month.
Growing Capabilities
Over the past year, Phyllis’s capabilities have continued to expand.
“When we went live with our new claims platform, Insurity, it brought us claim numbers in a completely different format,” said Zimmerman. “Our new numbers look like this: 225-000123456. Phyllis learned this new format when we went live in January 2024. However, we started to notice that we were receiving mail with the ‘dash’ removed. They would send us ‘225000123456.’ Without the dash, Phyllis was not matching/finding the claim numbers, which resulted in more work for us to manually review.” Roots helped to retrain Phyllis, and she learned to recognize the number and re-insert the dash.
Most recently, in December 2024, Phyllis learned to index high-priority mail. “It was the week of Christmas and was a great present for the team!” said Zimmerman. They had struggled with handling 125+ high-priority tasks that had to be indexed on the same day. This proved to be a challenge, especially on Mondays if they had a team member out of the office. Now they can easily work the high-priority mail with Phyllis each day.
Impressive Results
The benefits are substantial. Year to date, Phyllis has fully automated 67% of documents with no human intervention needed. She was able to reprocess and automate another 22%. Only 11% of her tasks have required manual processing.
Prior to Phyllis, average turnaround time for Claims mail was always three days. Now we have same-day processing.
The efficiency gains are impressive as well. Phyllis reduced processing time for “priority-five” mail from five days to just one hour—a 100-fold improvement. Overall, Eastern has saved more than 2,700 human hours since Q1 2023.
Most importantly, however, is that no humans were harmed when implementing Phyllis. She may do the work of several people, but she doesn’t replace our valuable human team members. “We have had no staff reductions with Phyllis,” said Harry Talbert, Eastern’s Senior Vice President, Information Systems. “The idea is to automate inefficiencies and reallocate people to other aspects of their work where they can provide more value—value that only a real person can provide.”
AI-powered digital coworkers are designed to streamline processes, not replace jobs.
Read more: “Meet Phyllis: Eastern’s AI-Powered Mailroom Associate,” ProVisions, February 2024, https://agents.proassurance.com/provisions/feb24#phyllis.
Your AI Library
In 2023, as part of our retention campaign, we offered the popular book Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again, by Dr. Eric Topol, cardiologist and executive vice president of Scripps Research.
In case you missed out, we have a few copies left. Let us know if you’re interested by emailing AskMarketing@ ProAssurance.com and include your mailing address.
About the Book
Topol examines the role AI will play in the future of medicine. He says that AI is quickly becoming an essential tool for cutting through the clutter to improve the speed and accuracy of diagnoses.
Deep Medicine provides an inspiring look at how deep learning algorithms can further a physician’s ability to create custom treatment plans. Topol has gathered a variety of examples of current AI use in medicine—from utilizing wearable sensor data to providing coaching via virtual assistants and reviewing medical histories via language processing. This book shows us how the power of AI can make medicine better and reveals the paradox that machine learning can actually make humans healthier—and even more human.
Video Resources
.jpg?width=300&name=sddefault%20(1).jpg)
Today's AI algorithms require tens of thousands of expensive medical images to detect a patient's disease. What if we could drastically reduce the amount of data needed to train an AI, making diagnoses low-cost and more effective? TED Fellow Pratik Shah is working on a clever system to do just that. Using an unorthodox AI approach, Shah has developed a technology that requires as few as 50 images to develop a working algorithm—and can even use photos taken on doctors’ cell phones to provide a diagnosis. In this TEDGlobal Talk, Shah shares how this new way to analyze medical information could lead to earlier detection of life-threatening illnesses and bring AI-assisted diagnosis to more healthcare settings worldwide.
In this TEDMED Talk, Greg Corrado discusses the enormous role that AI and machine learning will play in the future of health and medicine, and why doctors and other healthcare professionals must play a central role in that revolution. From distilling data insights to improving the decision-making process, Corrado sees a multitude of ways that AI and machine learning can help magnify the healing powers of doctors.
Listen as Dr. Siddhartha Mukherjee, Pulitzer Prize winning author of The Emperor of All Maladies interviews Dr. Eric Topol about his book, Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again.
In this TEDx Talk, Dr. Kurt Zatloukal, professor of pathology at the Medical University of Graz, Austria, and head of the Diagnostic and Research Center for Molecular Biomedicine, makes the case for why medicine needs AI.
Chatbot Cheat Sheet
Chatbots are AI solutions that simulate human-like conversations. If you have ever used your phone to ask Siri a question or asked Alexa to play your favorite song, you’ve interacted with a chatbot.
The New York Times compiled a list of phrases and concepts useful to understanding the new breed of AI-enabled chatbots like ChatGPT.
Bias
A type of error that can occur in a large language model if its output is skewed by the model’s training data. For example, a model may associate specific traits or professions with a certain race or gender, leading to inaccurate predictions and offensive responses.
Emergent Behavior
Unexpected or unintended abilities in a large language model, enabled by the model’s learning patterns and rules from its training data. For example, models that are trained on programming and coding sites can write new code. Other examples include creative abilities like composing poetry, music and fictional stories.
Hallucination
A well-known phenomenon in large language models, in which the system provides an answer that is factually incorrect, irrelevant or nonsensical, because of limitations in its training data and architecture.
For more on learning about AI, check out The New York Times’s five-part series on becoming an expert on chatbots.
Source
“Artificial Intelligence Glossary: Neural Networks and Other Terms Explained,” The New York Times, March 27, 2023, https://www.nytimes.com/article/ai-artificial-intelligence-glossary.html. (free subscription required)
.png?width=300&name=MicrosoftTeams-image%20(28).png)
The AI Paradox: Using AI to Discuss AI
While having dinner with my cousin Jessica, a dermatologist, I asked for her perspective on using artificial intelligence (AI) to screen patients. She responded dogmatically: “My attending said that AI will never be able to diagnose as well as the experienced eye of a physician.”
I challenged her assertion with some quick math: “How many skin lesions will a dermatologist see in their professional lifetime? 500,000? A million? AI doesn’t just 'see' a lesion; it breaks down thousands of data points per image. So, if an AI tool is trained on one million lesions, and each lesion produces hundreds to thousands of data points, you’re looking at billions of data points. Do you know any dermatologists who can do that?” She laughed and said, “Right ... tell the lawyers.”
The Invisible AI Revolution
Here’s the irony of modern healthcare: Many physicians remain openly skeptical about artificial intelligence while unknowingly using it every day. From a hospital’s EHR system that may include Clinical Decision Support (CDS) to that “smart” ventilator in the ICU, AI algorithms are already embedded in healthcare infrastructure—often with little transparency about who bears liability when things go wrong. While they’re enhancing patient care, they also raise new liability questions:
- Who’s responsible if the AI miscalculates?
- Did the clinician override the AI appropriately?
- Is the institution monitoring AI behavior like human staff?
As an MPL insurance agent, clients expect you to navigate this complex landscape and provide clear guidance. The challenge? How do you prepare for business conversations about AI when the field is evolving faster than any one human can track?
Your Secret Weapon: Using AI to Discuss AI
To calm my cousin’s knee-jerk rejection of AI’s utility, I reached for my iPhone and asked ChatGPT generative AI to list the liability risks associated with AI-aided diagnosis. In seconds, it returned a list that included diagnostic error liability, informed consent issues, standards of care impact, and documentation requirements. Jessica again laughed and said, “So now you’re using the very thing that’s causing liability issues to explain the liability issues?”
Here’s where the paradox becomes your advantage: You can use AI tools like ChatGPT, Google Gemini, or Microsoft 365 Copilot to prepare for sales calls and business meetings about AI:
- Before meeting with a client, use AI to summarize complex use cases specific to their specialty. Try prompts like: “Summarize the liability risks of using AI-powered predictive analytics in the emergency department.”
- Ask the AI tool to provide you with possible questions to use during the conversation based on those liability risks from the first step: “What questions might I ask a radiologist to develop a conversation about professional liability when using AI-assisted imaging?”
- Translate legal jargon into clinical reality: “Rephrase this AI exclusion clause in terms a cardiologist would relate to.”
- Roleplay responding to tough questions you might be asked: “Play the role of a neurologist concerned about using AI to analyze brain scans. Ask me challenging questions about liability and coverage.”
A word of caution: While AI can help you prepare, it shouldn’t replace your judgment or experience. Always review the AI’s responses, verify all facts, and bring your humanity to the conversation.
A New Opportunity
While some HCPs, like my cousin Jessica, remain skeptical about incorporating AI into their practice, the liability risks remain. That’s where you come in. Your clients need you to help guide them through the paradox where technology creates both the risk and the means to understand it. Master the balance between technological tools and human judgment, combine it with ProAssurance’s vast resources, and you’ll position yourself as the indispensable advisor your clients trust to help them navigate this complex landscape.
|
Written by Mace Horoff of Medical Sales Performance. Mace Horoff is a representative of Sales Pilot. He helps sales teams and individual representatives who sell medical devices, pharmaceuticals, biotechnology, healthcare services, and other healthcare-related products to sell more and earn more by employing a specialized healthcare system. Have a topic you’d like to see covered? Email your suggestions to AskMarketing@ProAssurance.com. |
Risk Management May Releases
With looming physician shortages and increased autonomy for certain Advanced Practice Providers (APPs), the relationships between supervisory physicians and APPs are more critical than ever before. Analysis of closed claims supports the idea that while individual members of this healthcare team maintain patient care responsibilities, certain risk reduction strategies aimed at enhanced communication, comprehensive documentation, and structured expectations increase patient safety, and improve quality of care and defensibility in the event of a claim.
Read More
The Allegation
The defendants failed to inform the mother of the risks of vaginal delivery with probable fetal macrosomia and failed to timely and correctly perform the maneuvers to relieve the shoulder dystocia.
Read MoreBrian Cools Has Retired
Brian Cools, Brand Strategist, retired this month. Brian served in various roles at ProAssurance over the past nine years, focusing on the development and management of the ProAssurance brand. Prior to his work at ProAssurance, Brian served as a trusted vendor doing marketing work for over 20 years.
Brian’s work can be seen on everything from banner stands and ProAssurance’s trade show booth to point-of-sale materials, the website and right here in ProVisions. Brian was a driving force behind converting what was once a simple monthly newsletter to a themed magazine.
Thank you, Brian, for all of your creative ideas through the years, and all of the fun along the way. We wish you well on all of your future endeavors.
- The ProVisions team
ProVisions Team
- Communications
- Design
- Digital Marketing
