Let’s slice pie from the sky and seriously consider AI’s role in workers’ compensation claims. First, admit that an end-to-end AI solution is impossible. WC does not provide a dispassionate linear process ripe for automation. Flow-charts or fishbone diagrams cannot codify WC claim contingencies. Rather, we must account for WC’s disparate interests and human unpredictability with a less analytical depiction… such as a pile of actual fish bones surrounded by alley cats fighting for scraps.
Forcing a vision of full-auto AI blinds us to the real possibilities. AI should not seek to master our status quo claim system. Rather, our existing system should radically change to best apply AI. Forget usual vendor roles and re-draw the process. With conventional models gone comes freedom to revolutionize.
I suggest we form logical “function modules” as powered by a sub-array of distinct AI tools, all recognizing boundaries for human connection and influence. Overseeing this flux of action is a super-adjuster in a role elevated from “data entry” to “data reaction.” New adjusters are extremely skilled and well-paid, acting as the human glue among a matrix of AI tools and able to react in real time to events and milestones in the interest of claim outcome all while managing the claimant experience.
As just a partial illustration, some newly defined function modules might include:
- Employee sentiment
- Reporting
- Work continuum
- Care direction
- Finances
- Closure
- Filings and compliance
- The “employee sentiment” module exemplifies just one new opportunity to support a known yet unmanaged challenge. In the current system, “employee advocacy” is a throw-away term with no defined responsibilities. “Employee sentiment” arises as a standard core concern with dedicated resources, focused purpose and ability to flag real-time urgencies. AI can support employee profiles before and during a claim. Continuous insight honed by machine learning ingests recorded statements, doctor findings, cooperative indicators, interim communications and non-claim-specific aspects. Human intervention is precisely triggered, gauged, placed where needed and not wasted. Nuanced automated messages can nudge desired reactions. More employees are better-cared-for while poorly motivated employees are dealt with from a smarter foundation. Less attorney representation is a critical result.
- The “reporting” module requires dedicated degrees of human interaction with nurses, investigators, employer contacts and intake staff. No “reporting” situation is fast-tracked because all information is valuable. This human investment returns value in securing fundamental data and spring-points for subsequent AI tools to act. As in old-school programs, good nurse triage can divert would-be claimants, yet we can add predictive analysis to round out a nurse’s conclusion.
- The “work continuum” module uses AI to evoke progressively updated, medically validated and predictively successful work opportunities with intensity far beyond common return to work (RTW) programs. Human interaction exists simply to coach a claimant among a wealth of opportunity. Beyond internal employer jobs are data-connected partnerships with out-placement temp agencies or non-profits, predictive validation for vocational rehab and labor market surveys. This module provides constant pressure and a holistic outlook. “Permanent restrictions” as a claimant legal tactic are conquered. A new employee culture is compelled to respect the ability to work.
- The “care direction” module replaces for-profit managed care with a dedication to predictively optimal care, with incentives provided by positive “employee sentiment.” AI validates treatment plans and choice of provider based on analytics matched to case-specific issues. Optimal healing is the paramount goal. Legislatures adjust laws concerning utilization review (UR) and customary care to allow weight for analytics as combined with traditional medical evidences, all with confidence in patient outcome over legal gamesmanship.
- The “finances” module might run mostly on AI, including basic management of employer funding schemes, premiums, claim payments and reserving. Medical bills are adjusted based on vast analytical data that supports any jurisdictional scheme. Percentage-of-savings fees no longer exist. Aggregate employer-population contribution from the “employee sentiment” module builds weighing factors to fortify premium calculations and actuarial outlooks.
- The “closure” module applies AI to provide a progressive outlook for method of resolution and closure date. Predictive likelihoods around dollar value and strategies to settle or litigate and even subrogation value are updated in real time. Data on judges, mediators, and lawyers feeds machine learning. The array of related AI tools herein includes analytically derived Medicare Set Asides (MSAs) and annuities.
- The “filings and compliance” module can be highly automated, fed first from the “reporting” module and updated from other modules in real time while running the gamut of statutory and OSHA requirements. Even this benign stockpile of data can be a reciprocal source of machine learning among other AI cells.
