- Key insight: How a workers comp insurer navigates AI agent tools
- Forward look: AI agents can help 'reach customers where they are'
Third-party AI agents, especially Meta's Muse, are
This article is excerpted from a longer set of responses and edited for clarity.
Does Employers allow Meta's Muse tool to interact with its operations?
We don't currently allow third-party AI agents like Muse to directly access our operational systems or customer data. Our approach is intentional: we've invested in building an enterprise-approved AI stack that we fully control and govern, which lets us move with both speed and confidence.
We're closely watching how the industry is managing these relationships. We've seen other companies, such as Amazon, set clear boundaries around data access and consent, and we think that's the right framework. If there were a partnership opportunity with a tool like Muse that came with proper alignment around data usage, governance standards, and transparency, we'd evaluate it. But it would need to work within our broader AI strategy, not replace it.
Does Employers use or allow access for any similar AI tools, such as ChatGPT?
Yes, but strategically. We built an application within ChatGPT to provide quoting insights to our agents and customers which is an example of how we're exploring AI tools in ways that create real value to potential customers while maintaining data governance. Additionally, our employees have access to enterprise-approved AI tools, but we don't allow personal accounts or unrestricted use of consumer-facing AI platforms. We provide access to frontier and open-source models through our internal AI infrastructure, which means we get innovation benefits while maintaining visibility and control. It's about being thoughtful, not being afraid. The tools are powerful; the key is knowing how and where to use them.
Do Muse and other AI tools conduct themselves in an acceptable manner? Do they produce incomplete or distorted information?
This is exactly why governance matters early. The insurance industry learned some hard lessons in the early days of AI deployment where there were instances where AI tools made commitments or claims that didn't align with actual policy or product, and companies had to correct the record. We've structured our approach around a few principles: clarity on data usage, audit visibility, and continuous monitoring. When we work with any external tool or vendor, we're clear about what data is being used, how it's being used, and what we expect in return.
The NAIC Model Bulletin, a regulatory framework created by state insurance regulators, reflects this thinking too: insurers remain accountable for AI systems in their ecosystem, whether built internally or provided externally. Prevention through clear governance is better than firefighting.
Are there benefits to allowing Muse or other AI tools to interact with your operations?
Absolutely. The potential is real. If a tool like Muse could help us reach customers where they already are, maybe they're asking their AI assistant about workers' comp options. That's genuinely valuable. It's about meeting customers in their workflow, not forcing them into ours. The opportunity is in frictionless access to information and services. The challenge is doing it responsibly. The smartest companies in the tech and financial services space are figuring out how to use AI platforms in ways that create mutual value, while protecting customer interests.
When you invest in AI governance early by thinking through policies, risk management, and vendor relationships before you're in crisis mode, you can move faster and with more confidence. Our ChatGPT application is a great example of this. We were able to build and deploy it quickly because we'd already worked through the governance questions: What are we using it for? What data is involved? What safeguards do we need?
We focus on applying AI to the pieces of the insurance workflow that matter most: faster claims decisions, better information for our adjusters, smoother interactions for agents and customers, and ultimately better service for injured workers.
