In workers compensation claims, outcomes have historically depended on the quality of experts — both the claims adjuster and, less visibly, a team lending support to adjusters behind the scenes. What if that team wasn’t just made up of human experts but also included a squad of artificial intelligence bots, or “agents,” working in tandem, with each ready to assist with its specialized skills? That’s the future, and it’s closer than you might think.

Generative AI is the buzzword of the day, grabbing headlines left and right. But for those in the know, AI has been a secret weapon for years. For forward-thinking organizations, it’s been a quiet partner in making decisions that significantly impact claim outcomes. 

From reserving to clinical intervention, provider selection to litigation avoidance, and subrogation to return to work, AI has been there, quietly guiding key decisions and improving outcomes. It’s the brainy assistant that never sleeps, tirelessly sifting through data to find golden nuggets of insight.

Up to this point, these various AI models have been spokes feeding information back to a human hub, typically the adjuster. As AI continues to proliferate, this could soon be akin to a basketball team in which the players only ever communicate with the coach — and never with each other. Whether on the court or with a claim, communication flows best when all the players on the team are actively engaging with one another.

The solution is to develop AI agents that can communicate with one another, working as a bona fide team and ultimately presenting human experts with a cohesive plan. That plan will support the claims adjuster with recommendations and supporting information in making the critical decisions on the claim — decisions which, for the foreseeable future, will remain in the expert hands of the human adjuster.

Notably, large AI companies like Google and OpenAI, makers of ChatGPT, use a similar approach in creating their AI models. These GPT models are refined using an approach in which one AI agent proposes a response to an input, and a second AI agent grades the response based on what it has learned about human preferences. Similarly, in the claims world two or more AI agents working can generate significantly better recommendations when they work together. 

To illustrate what this could look like in workers compensation, start by imagining components of your dream team for managing claims: intake and assignment staff to get things rolling; clinical oversight to flag clinical/psychosocial risks and mitigate them; litigation avoidance and fraud investigator experts to navigate and mitigate risks; a claim auditor to review output and ensure consistently high quality across the team; a comp law expert to provide up-to-date legal insights as needed; a licensed adjuster, accountable for critical claim decisions; and an administrative assistant to help coordinate the team.

Now, let’s leap into the future where this team is AI-based. The moment a claim is filed, the AI team begins its work. An AI intake agent takes the First Notice of Loss and follows up with relevant questions helpful to the investigation of the claim. Key pieces of information are passed to AI teammates for further processing. Then, the clinical oversight agent reviews the information passed to it from the AI intake agent and spots signs of acute psychosocial challenges. It recommends immediate human clinical intervention and sends an alert to the litigation avoidance agent, given the observed relationships between psychosocial risks and litigation. Later, the claim auditor agent reviews claim facts collected to date and notices factual inconsistencies in verbal and written correspondence provided from intake. It engages the fraud investigator agent with this information to gauge the risk of fraud and determine appropriate next steps to mitigate risk. The legal expert reviews action plans recommended by the AI team and ensures adherence to relevant laws. Any regulations that would require action by the adjuster are flagged for follow-up. 

Then, a claim assignment agent reviews output passed from the agents above, which it uses to determine the claim’s unique risks and complexity. Based on this, it identifies the optimal licensed adjuster to handle the claim. Lastly, the administrative assistant agent compiles all findings, including action items, and schedules time for the designated adjuster to review them the next day.

By the time the human adjuster logs in, a coherent and cohesive plan is waiting for them. This AI dream team stays engaged 24/7, offering unwavering support throughout the claim’s lifecycle.

This is the future of claims management: a blend of AI efficiency and human expertise, where much of the heavy lifting is done before the sun even rises. It’s a vision in which AI agents work together to provide and empower the human adjuster with the information required to make the nuanced, critical decisions that truly require human judgment. 

The technology to create a seamless, efficient and thorough claims management process is largely already here. However, it demands a paradigm shift — from isolated AI solutions to holistic AI teams that can collaborate effectively. 

Currently, the market is flooded with AI solutions, but they often leave the human professional as the central “hub” in the wheel among an ever-growing number of AI “spokes,” potentially creating a choke-point as humans coordinate with disconnected AI models. What we need is an architecture in which the AI agents can self-coordinate.

There are two ways to capitalize on this opportunity. You can architect your own AI team. To build this team, you need to be great at not just building AI agents but also getting them to work as a team. One key to success in building your agents is establishing a rigorous evaluation methodology, ensuring your agents consistently outperform humans at the same task. Getting the agents to work as a team requires a long-term vision for the team’s composition and rapid iteration to make that vision a reality. Or you can partner with a group with this vision. This involves partnering with firms with a vision for a team of AI agents. Be wary of firms that offer a collection of independent AI models with no plans to integrate these into a team. 

Whether you want to build your own AI team or partner with a visionary group, the key is to be aware of the promise and peril of AI teamwork, and how they can be addressed with superior claim outcomes in mind. AI agents can achieve more and have more profound claims impact together than alone, but they need careful design, evaluation and integration to perform well. By planning for this opportunity, you can gain a competitive edge.

Joe Powell is senior vice president of analytics and product innovation at Gallagher Bassett. He can be reached at joseph_powell@gbtpa.com.