Overview
What is Artificial Intelligence?
Artificial Intelligence (“AI”) is a technology that enables computer systems to complete tasks that normally require a human’s intelligent behavior, such as collecting information, analyzing data, and making decisions. As it continues to improve, AI has become more proficient at tasks that have historically been difficult for computers, such as recognizing images, identifying spoken words, and analyzing unstructured data.[i] As such, it can be used to recognize patterns and predict future data. AI advancements continue to increase and affect multiple industries, including insurance claims.
Why AI is Relevant Today
AI’s use has risen dramatically because of rapidly evolving technology, and in particular, the increase in accessible data, greater computing capabilities and changing consumer expectations. “Higher computing power, memory capacity, cloud computing, big data technologies, and global connectivity have enabled machines to run complex algorithms faster and handle more data than a human could.”[ii] As a result, numerous applications and uses of AI technology have emerged across all industries. In the insurance industry, AI has gained popularity in multiple areas, including claims handling, underwriting, risk evaluation and fraud detection.[iii] In June of 2023, Lemonade Insurance, which emphasizes its use of AI and machine learning technology, reportedly broke a world record by settling an insurance claim within two seconds by using AI and machine learning. [iv] However, AI has also been in the news for less favorable reasons. For example, also in June of 2023, after multiple proceedings over several months, a federal judge for the U.S. District Court, Southern District of New York issued a 34 page Opinion and Order of Sanctions, after an attorney used the AI tool, ChatGPT, to prepare a pleading that cited and quoted from purported decisions said to be published in the Federal Reporter, the Federal Supplement and Westlaw, but which in reality, had been fabricated by ChatGPT.[v] Courts across the country have warned or sanctioned attorneys for their failure to address these AI “hallucinations”. As recently as February of this year the U.S. District Court, District of Wyoming sanctioned attorneys with the law firm Morgan & Morgan for their failure to confirm that cases cited in AI-drafted court submissions were real, emphasizing an attorney’s duty to “make a reasonable inquiry into the law before signing . . . a document.”
Insurers have also been challenged over their use of AI. In a class action lawsuit pending in the U.S. District Court, District of Minnesota, plaintiffs allege that certain Medicare Advantage insurers used AI in lieu of physicians to make coverage determinations and in doing so ignored the recommendations of the plaintiffs’ treating physicians. As this and others are the earliest cases on these issues, their outcomes may shape the future of regulations and industry standards on the use of AI in insurance.
Benefits of AI for Insurance
Improved Support for Decision Making
Improved support for decision making in the claims process is often cited as a benefit of AI. A vital part of claims handing usually involves data gathering and analysis. AI applications can be used to perform these tasks quickly and effectively. During this process, AI can also be used for fraud detection. A 2021 survey conducted on behalf of the Coalition Against Insurance Fraud determined that insurance companies are increasingly adopting technologies such as AI for detection of claims fraud.[vi] Key components of such fraud detection included automated red flags, predictive modeling, text mining, reporting capability, case management, exception reporting, and data visualization/link analysis.[vii] Finally, the survey discovered that detection of claims fraud was the highest area of technology that the companies were considering investing in in the immediate future.[viii]
Increased Productivity
Companies have found that AI can increase productivity. Adjusters can use AI applications, such as claims management software, to streamline the claims processes, including initial data scanning and processing and verification of policy details and identification of gaps and errors.[ix] This clears up time for the adjusters to focus on other tasks.[x] AI also provides for more efficient claims routing and is particularly beneficial in the resolution of more routine claims, such as the one resolved by Lemonade.
Reduced Costs
As could be expected from the benefits report above, another benefit of AI can be reduced costs. It is reported that AI can lead to $390 billion in cost savings for insurers by 2023.[xi] For example, AI can be used to quickly digest large amounts of information submitted by claimants and insureds and enable human claim staff to spend less time analyzing irrelevant information, and more time analyzing the information that will be critical in resolving the claim.
Enhanced Customer Experience
A 2021 AI Insurance Survey by PwC reported that the most effective use of AI in insurance companies is in customer experience.[xii] AI can immediately confirm the existence of a policy and provide information to the claims adjuster concerning the policy term, limits and deductibles. AI should also be able to provide loss runs for brokers and insureds. AI can immediately route calls to the appropriate contact person at the insurer. Similarly, AI has the potential to streamline the renewal process and provide guidance on the potential rate range.
Pitfalls of AI for Insurers
Concern for Human Review and Analysis
AI is known for it’s ‘hallucinations’, which warrants meticulous review by an individual user. Despite this risk, for many aspects of claims handling, AI can strengthen the process and help to ensure that a claim is adjusted quickly and fairly. For example, when a fifty-page complaint is filed against an insured, AI can digest the complaint and summarize the allegations and make it easier for the claim adjuster to get started toward helping the insured to resolve the claim. Similarly, when a claimant provides hundreds of pages of medical records, AI can summarize them quickly so the claims adjuster can understand the full scope of the claimant’s injuries and their relationship to the occurrence. Tokio Marine incorporated an AI-based claims document upload system to upload handwritten notices into its system, which cut input time by 50%, cut human error by 80% and fast-tracked claims payments.[xiii] Insurers have also begun to use AI to detect fraudulent claims activity.[xiv]
Aside from the technical challenges of dealing with AI, such as ensuring that the data set used by the AI system is large enough to be valid, claims administration will continue to require human review and analysis. For example, in third party claims, the adjuster must determine whether there is a duty to defend, who should defend the insured, and then must determine a fair settlement value for the claim, or whether the case should be tried. If a settlement demand is received, the adjuster has to consider the likelihood of liability and the likely damages that would be awarded if the case proceeds to trial. This requires an analysis of the legal positions of each party, the facts learned in discovery, and a comparison between those facts and jury verdicts in similar cases in similar jurisdictions. While much of this can be assisted by AI, it will be essential for insurers to have humans make any final judgments about claim payments, particularly if the insurer decides not to settle the claim or decides that there is either no duty to defend or indemnify the insured.
Every state has rules that require insurers to investigate claims fully and fairly, and the failure to fully and fairly investigate a claim is likely to lead to a bad faith suit, which may involve statutory penalties, compensatory damages and exemplary damages.[xv] To ensure that claims are handled according to the standards of good faith and fair dealing, the insurer must be able to document that it fully and fairly investigated the claims and that it reasonably informed the insured claimant of the basis for its decision. Typically, this is established by the claim notes entered by claims handlers in the file. Once a bad faith suit is filed against the insurer, the insurer’s good faith will be established by the testimony of the insurer’s claims staff.
The potential for such suits is not just theoretical. In July 2023, Cigna was sued in California, and the plaintiff alleged that Cigna “unlawfully reviewed insurance claims using automated systems rather than relying on humans.”[xvi] The Complaint relied on California regulations requiring every insurer to conduct and diligently pursue, through a fair and objective investigation information necessary to resolve the complaint.[xvii] The suit claimed that Cigna’s claim reviews were not thorough, fair, or objective because the company made claim determinations without conducting any investigation, simply on the basis of its computer algorithms, and without the review of the medical director.[xviii] Not surprisingly, Cigna has denied that its process violates California regulations, and denies that the claim denials are based solely on the computer process.
To the extent decisions have been based on AI, without substantial input from the humans on the claim staff, the insurer will need to demonstrate how the AI came up with the decision, which will mean explaining to a judge or jury the algorithms used by the AI, the source of the data used by the AI, and the protocols used by the insurer to ensure that the result was fair to the insured. To the extent that humans on the claim staff were also involved in the decision and analyzed the result to determine whether it was fair and complete, the case will be easier to defend, provided that the claim staff fully documented their decisions and the bases for those decisions.
Prejudice/Bias
One area of concern for insurers using AI in any area of their business is the potential for bias in the algorithms underlying the AI. To the extent that these algorithms are biased, or based on misleading or incomplete data, it can lead to unfair discrimination against insureds.[xix] The elimination of bias and prejudice from the algorithms will require insurers to closely examine the databases used for the AI. As one author explained, to address AI bias, “one must turn to the source: the collection of actuarial data and the way employees interact with it.”[xx] This will also be true in the collection of claims data needed to enable AI to be useful in the claims process.
Data that goes into an AI database is collected over years, and any biases in the data will be deeply ingrained.[xxi] Similarly, as humans input the data and select which data sets to use, their biases will be incorporated into the data.[xxii] As a result, it will be essential that, as AI algorithms are created and used, they be reviewed and analyzed to ensure that bias is eliminated or at least minimized.
Privacy Concerns
Another issue posed by AI is privacy. There are many areas of the claims process where privacy and controlling access to information is essential. For example, in any bodily injury claim, the insurer will be provided with the claimant’s medical records, which are subject to the privacy requirements of the HIPAA privacy rule.[xxiii] In third party claims where the insurer defends the insured, information provided by the insured will typically be subject to the attorney client privilege.[xxiv]
The extent to which privacy can be maintained will depend on the manner in which the AI system is created. For example, the use of AI generation software such as ChatGPT prevents any expectation of privacy because searches become part of the database and part of the public domain. Needless to say, public AI sources like ChatGPT should not be used for claim handling simply because everything will lose any privacy protection.
If an insurer develops its own AI system, it likely can ensure that the data used remains within the insurer’s possession by limiting access to the insurer’s employees who need to have access. When an insurer uses a third party’s AI system, it will need to be careful to prevent a loss of privacy. This will require carefully vetting the third-party AI vendor to ensure that it has adequate protocols to protect data privacy. The insurer should also include confidentiality requirements in its contracts with these vendors that require the vendor to maintain confidentiality and limit access to the insurer’s data to those who need to have such access.
Regulatory/Legislative Concerns
Nearly every aspect of insurance is subject to statutory and regulatory guidelines. While most state legislatures have not started to address the use of AI in insurance, the National Association of Insurance Commissioners (NAIC) approved in December 2023 its guidance entitled “NAIC Model Bulletin: Use of Algorithms, Predictive Models, and Artificial Intelligence Systems by Insurers.” Since that time, approximately half the states insurance departments as well as D.C. have adopted a version of the bulletin, while others, like New York (and its Insurance Circular Letter No. 7), have issued alternative guidance when it comes to AI use in insurance. The NAIC bulletin encourages each state insurance department to remind insurers authorized to do business in the state that “decisions impacting consumers that are made or supported by advanced analytical and computational technologies, including artificial intelligence (AI) systems …, must comply with all applicable insurance laws and regulations.”[xxv] In order to mitigate and minimize adverse consumer outcomes, the guidance encourages the development of an AIS program and strong corporate governance regarding the use of AI, building AI into an enterprise risk management framework, and overseeing third party vendors who use AI models adopted by the insurance industry, among other things. Additionally, in 2020 the NAIC adopted the NAIC Principles on Artificial Intelligence, setting forth the expectation that insurers who adopt AI should ensure that the AI systems used are fair and ethical, accountable, compliant with applicable insurance laws and regulations, transparent, secure, safe and robust.[xxvi] As AI becomes more prevalent in the insurance industry, state insurance departments are likely to seek to regulate the process more completely.
Good Faith Use of AI for Claims Handling
Applications for AI in Claims Processes
There is little question that AI has a place in the claims process, as it can help insurers streamline the claims process and quickly digest large amounts of information. The earliest uses are likely to be in summarizing information obtained from insureds, claimants and others so that the claims handlers can more readily understand the various issues presented. AI can be used to summarize medical records, depositions, witness statements and other voluminous materials. Further, AI can be used to assist with generating email correspondence, as well as letters and related insurance correspondence. Similarly, to the extent the data is available, AI can be used to analyze jury verdicts in various jurisdictions for cases with similar facts to predict likely outcomes and settlement values of cases. AI can also be used to quickly adjust uncomplicated claims, which will likely enhance consumer satisfaction with the claims process.
Establishing Policies and Procedures for Use of AI
As insurers incorporate AI into their claims processes, it will be important for them to develop policies and procedures to minimize or eliminate bias and ensure oversight of claims decisions by experienced claims personnel. This will mean that the programs must be reviewed to prevent arbitrary decisions and to make sure that the data being used, and algorithms are not biased.
It will be important as AI is adopted for the insurer to be able to explain how its algorithms were created and the steps that were used to ensure the accuracy of the data being used and how they were checked to ensure that bias has been minimized. Also, insurers should establish procedures to review the output of the AI and verify that it is accurate and unbiased.
Training and Education for Claims Handlers
As AI is adopted, insurers will need to provide training to claims handlers on the use AI in claims decisions. They will need to understand the limits of AI and the need to use their own judgment and experience to verify that AI output complies with the legal and ethical standards for fair claims handling. They need to understand that AI is a tool for their use, but that it cannot replace their own judgment concerning the way the claim should be adjusted.
Future Outlook for AI in Claims Handling
There is little question that AI is here to stay in all aspects of insurance, including claims handling. For now, insurers need to take care to ensure that adoption of AI enhances the experience of insureds and claimants, without making claims decisions that are adverse to the insured in the absence of human review and analysis.
It seems unlikely that AI will be ready to perform the type of analysis needed to resolve complex claims in the next few years. However, AI is already able to streamline the claims process, particularly for routine types of claims.
[i] NAIC, July 23, 2023, https://content.naic.org/ cipr_topics/topic_artificial_intelligence.html.
[ii] Id.
[iii] See Antony Xavier, 6 Ways Machine Learning and AI are Transforming the Insurance Industry, SIMPLESOLVE INC., March 12, 2021, https://www.simplesolve.com/ blog/machine-learning-and-ai-in-insurance-industry/ (last visited Apr 27, 2025).
[iv] See Jack Willard, Lemonade shatters record by using AI to settle a claim in two seconds, Reinsurance News, June 14, 2023, https://www.reinsurancene.ws/lemonade-shatters-record-by-using-ai-to-settle-a-claim-in-two-seconds/ (last visited Apr 27, 2025).
[v] Mata v. Avianca, Inc., 2023 WL 4114965, at *17 (S.D.N.Y. 2023)2023 WL 4114965, at *17 (S.D.N.Y. 2023)
[v] Wadsworth v. Walmart, Inc., 2025 WL 608073, at *5 (D.Wyo. 2025).
[v] Plaintiffs’ Class Action Complaint at 1, Estate of Gene B. Lokken, et al. v. United Health Group, Inc., et al., Docket No. 0:23-cv-03514 (D.Minn. 2023).
[vi] See SAS Coalition Against Insurance Fraud, State of Insurance Fraud Technology Study, 2021 Results, https://www.sas.com/content/dam/SAS/documents/marketing-whitepapers-ebooks/third-party-whitepapers/en/coalition-against-insurance-fraud-the-state-of-insurance-fraud-technology-105976.pdf (last visited Apr 27, 2025).
[vii] Id.
[viii] Id.
[ix] See Xavier, supra.
[x] Id.
[xi] Id.
[xii] See PwC 2021 AI Insurance Survey, https://www.pwc.com/us/en/tech-effect/ai-analytics/ai-predictions/insurance.html.
[xiii] Anthony Xavier, 6 Ways Machine Learning and AI are Transforming the Insurance Industry, Simple Solve Inc., March 12, 2021, https://www.simplesolve.com/blog/machine-learning-and-ai-in-insurance-industry/ (last visited Apr 27, 2025).
[xiv] See Jim Hulett, Artificial intelligence in claims fraud: How AI is automating suspicious behavior detection, Verisk, July 15, 2020, https://www.verisk.com/insurance/visualize/artificial-intelligence-in-claims-fraud-how-ai-is-automating-suspicious-behavior-detection/ (last visited Apr 27, 2025).
[xv] E.g., Harner v. USAA General Indemnity Co., 497 F.Supp.3d 901, 911 (S.D.Cal. 2020)(delayed payments because of inadequate or tardy investigations and oppressive conduct by claims adjusters may breach the implied covenant of good faith and fair dealing.); Bailey v. Farmers Ins. Co., Inc., 137 P.3d 1260, 1264 (Okla.Civ.App.2006) (claim for breach of duty of good faith and fair dealing requires showing that the insurer’s refusal to pay was unreasonable under the circumstances because it had no reasonable basis for the refusal, it did not perform a proper investigation, or it did not evaluate the results of the investigation properly.)
[xvi] T/ Claburn, Cigna sued for using software to deny healthcare insurance claims, The Register, July 26, 2023, https://www.theregister.com/2023/07/26/cigna_healthcare_insurance_ai/ (last visited Apr 27, 2025).
[xvii] Id.
[xviii] Id.
[xix] F. Huang, How insurers can mitigate the discrimination risk posed b AI, USNW, BusinessThink, July 13, 2022, https://www.businessthink.unsw.edu.au/articles/insurance-ai-discrimination (last visited Apr 27, 2025).
[xx] Ifty Kerzner, Marketing Data Can Reduce AI Bias in the Insurance Industry, Insurance Journal, Feb. 20, 2023, https://www.insurancejournal.com/magazines/mag-features/2023/02/20/708144.htm (last visited Apr 27, 2025).
[xxi] Id.
[xxii] Id.
[xxiii] 45 CFR Parts 160 and 164.
[xxiv] E.g., Chicago Trust Co. v. Cook County Hosp., 298 Ill.App.3d 396, 407, 698 N.E.2d 641 (Ill. Ct. App. 1st Dist. 1998).
[xxv] NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers, Adopted 12/2/2023, available at https://content.naic.org/sites/default/files/2023-12-4%252520Model%252520Bulletin_Adopted_0.pdf (last visited Apr 27, 2025).
[xxvi] National Association of Insurance Commissioners (NAIC) Principles on Artificial Intelligence (AI), Adopted 8/14/2020,available at https://content.naic.org/sites/default/files/inline-files/NAIC%20Principles%20on%20AI.pdf.(last visited April 27, 2025).