Insureds’ growing use of generative artificial intelligence to submit claims and contest adverse determinations is challenging insurers’ claims management systems and protocols. For insurers, this increased use of gen AI by insureds creates a potential disruption to the traditional claim handling process in addition to increasing overall costs and workload.

The challenge for insurers is to develop better workflow strategies without allowing the higher complexity or appearance of the paperwork to turn into higher loss adjustment expenses. When developing better workstream flows, claims managers must assess:

• Who owns the response.
• What warrants additional specialist involvement.
• If the company’s own AI tools are proper and being properly utilized.
• If legal resources are being utilized in the right kinds of cases.
• How leadership can assess whether claims are being allocated to the right resources.

Adjusting to the Influx of Pro Se Litigation

Insurers are seeing more insureds taking a run at representing themselves in litigation over adverse claims determinations. Beyond anecdotal observation, reported federal litigation research provides context. Research published earlier this year by Anand Shah of the Massachusetts Institute of Technology and Joshua Levy of the University of Southern California found an annual average of approximately 23,000 non-prisoner pro se federal lawsuits between fiscal years 2005 and 2022, followed by 27,370 in FY 2023, 31,478 in FY 2024 and 41,490 in FY 2025. Their analysis also found that non-prisoner self-represented litigants accounted for about 11% of federal civil filings during the earlier period, compared with 14.5% in FY 2024 and 16.8% in FY 2025.

The researchers also examined a random sample of 1,600 federal civil complaints, comprising 200 per year from 2019 through early 2026. They estimated that the share containing AI-generated text rose from 0.1% during 2019-2022 to 3.5% in 2024, 10.5% in 2025 and 18% in early 2026. These are estimates of AI-generated text, not confirmed tool-use records.

Separate research by Or Cohen-Sasson, director of the Miami Law & AI Lab at the University of Miami School of Law, found that non-prisoner pro se litigation represented 11.3% of federal civil filings before 2022 and 22.3% by the final quarter of FY 2025. That quarterly comparison should not be treated as interchangeable with Shah and Levy’s annual figures.

These findings neither establish that AI caused the increase nor measure claim frequency or insurance defense spending. We nevertheless see more self-represented litigants pursuing cases that traditional counsel may decline. They increasingly use AI to draft pleadings, navigate litigation procedures, develop extensive discovery, and prepare motions and responses. Carriers should monitor how these trends increase the work and costs of assessing and resolving disputes.

Separate AI-Enhanced Presentation From Substance

The reported outcome findings reinforce the need to distinguish presentation from merit. Shah and Levy found that pro se dispositions remained broadly stable. Dismissals moved from approximately 60% before widespread generative AI use to 63% afterward, settlements from roughly 20% to 23%, and self-represented plaintiffs continued to prevail in fewer than 1% of cases in their analysis.

Cohen-Sasson’s study reported dismissal rates of 61.1% for complaints flagged as likely to contain AI-generated text, compared with 53.6% for unflagged complaints. It also reported that 47.8% of flagged cases ended before the defendant answered, compared with 39% of unflagged cases. These are descriptive comparisons, not proof that AI caused the outcomes. The two studies should not be treated as measuring identical populations.

A well-organized demand can make a genuine problem easier to understand. It can also make unsupported assertions appear more persuasive. Claims teams should resist both the temptation to equate professional presentation with substantive merit and the opposite temptation to dismiss a submission because it appears AI-assisted. The starting point should be the same regardless of drafting method. What happened, what evidence supports the position, and what decision is the carrier being asked to reconsider?

That requires distinguishing factual assertions from legal conclusions and separating new information from repeated arguments. A submission may invoke several legal theories while raising only one unresolved factual issue. Another may contain little legal terminology but identify an important missing record or an error in the carrier’s evaluation.

Staff training should focus on those distinctions. Trying to identify whether a document was written by AI is less useful than identifying whether its material assertions are supported.

Redesign Processes for AI-Enhanced Submissions

Carriers are seeing claimants use AI to provide several revised demands with increasingly elaborate arguments but no new information that actually impacts the claim analysis. Using technology and developing forms to help the claims professional quickly identify what has already been addressed and what still requires a response can be key.

Ownership of each stage of the claim process should be explicit and coordinated, taking the increased complexity caused by AI use into account. As always, the line claims professional’s authority needs to be properly matched to training and ability, and supervisors should address matters outside any delegated authority. Thought should be given to whether and to what extent AI-enhanced submissions merit retailoring where the authority lines are set—for example, when should claims be sent for separate legal analysis to counsel in the insurer’s law department or when should they be sent to outside counsel, etc. Traditional internal rules may need to be adjusted in light of the AI enhancements.

Give Legal Departments a Defined Role

AI-assisted claims correspondence often presents legal issues that require legal expertise before litigation begins. For example, unfamiliar authorities, disputed policy interpretations, threatened extracontractual claims or questions about the carrier’s response obligations should all be reviewed and analyzed by appropriately trained personnel. We see this creating much more work but also causing stress on internal legal teams for many insurers that routinely send claims with allegedly supporting legal arguments or bad-faith rhetoric to the legal or law departments.

When matters need to be assessed by the legal department should be thoughtfully delineated. Assessing how increased workflows may impact the legal infrastructure should be recognized and addressed by management. Carriers should look for strategies for addressing these new enhanced demands without making legal departments the default reviewer for every lengthy letter. Escalation criteria should reflect legal significance, deadlines and potential prejudice, not legalese and document length alone.

Focus Outside Counsel’s Assignments

Similar to developing processes for assessing when claims professionals should refer matters to internal counsel, companies should consider how increased AI-enhanced work may need to be sent to panel counsel or outside counsel. A submission’s numerous arguments do not necessarily require separate, extended analyses of every point. Counsel and the carrier should identify which issues affect the outcome and which can be addressed together. Material inaccuracies should be addressed directly, but a response to every error can create expense without advancing resolution. Litigation strategy and budgeting must reflect the governing rules and the case’s circumstances. Suspected AI use alone does not necessarily dictate the strategy, but the additional workload may impact budgeting and work allocations.

Measure the Burden Instead of Assuming It

Claims leaders should test whether changing submission patterns are actually creating bottlenecks. Useful measures include time spent reviewing repeat demands, legal referrals, specialist costs, authority-verification work, defense spending and time to resolution. Those measures should distinguish new evidence from repetitive positions and be considered alongside claim outcomes. A referral that identifies a genuine evaluation error is not necessarily wasted work. Conversely, repeated review by several departments may reveal an unclear assignment of responsibility.

The carrier does not need to identify every AI-assisted document to improve its process. It needs to know where work is accumulating, why specialists are being engaged, and whether those interventions help resolve the dispute.

AI may make it easier to present a claim without making it easier to resolve one. Insurers cannot control the drafting tools available to insureds, claimants or litigants, but they can control the discipline of their response. Carriers should identify the issue, verify material assertions, assign an owner and engage expertise when it is needed.