AI is everywhere. It is in boardroom conversations, vendor demos, conference sessions, LinkedIn feeds, and even Netflix commercial breaks. For insurance carriers, especially claims teams, the pressure to “do something with AI” is only growing.
But as Jim Girard, VP of Engineering at Benekiva, shared in a recent Claimversation, the question is not simply, “How can we use AI?”
The better question is: Where can AI actually help, and how do we use it responsibly?
Jim has spent more than 30 years in insurance technology, including more than 25 years focused specifically on the insurance industry. He is also clear about one thing:
“I’m pro-AI,” Jim shared. “I’m all in it.”
He uses AI regularly, from finding information to expanding ideas and supporting creativity. He sees the potential, both personally and professionally. But when that potential is brought into the insurance and claims environment, the conversation becomes more complex.
Because claims is not just a workflow. Claims is where technology, process, regulation, cost, security, and human care all meet. Listen to the full conversation here.
One of Jim’s strongest recommendations for carriers is to take a step back before jumping into AI. Look at the process first. Many claims processes were designed years ago, long before today’s technology was available. So before asking where AI fits, carriers should ask whether the process itself still makes sense.
From there, Jim recommends looking for the “mundane” points in the process: the repetitive, time-consuming, necessary tasks that slow people down but do not necessarily require human judgment every step of the way. These may include document processing, summarization, correspondence support, research, or other administrative steps that consume examiner time.
The goal is not to add AI for the sake of adding AI. The goal is to identify where technology can reduce friction. Then, and only then, carriers can ask: Can AI make this stronger?
That distinction matters. AI should not be treated as a strategy by itself. It should be evaluated as a tool that may strengthen an already thoughtful process.
A key theme from the conversation was the difference between being AI-assisted and AI-driven.
AI-assisted means using AI to support people. It helps with the work around the decision: gathering information, summarizing documents, identifying next steps, drafting communications, or helping examiners move through claims more efficiently.
AI-driven means relying on AI to own the process end to end. For Jim, that distinction is especially important in insurance. Insurance is highly regulated. Claims decisions can have legal, financial, and deeply personal consequences. If something goes wrong, a carrier cannot simply say, “The AI made the decision.”
That is why human oversight matters, the industry concept of “human in the loop.” AI can help examiners move faster and spend less time on monotonous work, but people still need to be present at the right checkpoints to ensure decisions are accurate, compliant, and care-filled.
Jim put it plainly: when AI becomes fully driven, there is a risk of taking the passion, care, and empathy out of the process. And in claims, that care matters. For many claimants, the interaction may happen during one of the worst days of their lives. Technology should help examiners better serve that person, not create more distance between them.
One of the most important reminders from the conversation was also one of the simplest:
Claims is part of the care industry.
Yes, claims teams process information. They review documents. They follow rules. They confirm eligibility. They move work forward. But at the center of it all is a person. Someone may be filing a death claim, a critical illness claim, or another claim tied to a difficult life event. That person is not thinking about workflow optimization or system architecture. They want to be understood. They want clarity. They want the process to work.
That is where AI has real potential.
Used thoughtfully, AI can help examiners understand the claimant’s situation more quickly. It can reduce the administrative burden that keeps teams buried in manual work. It can support better, faster, more informed service. But AI should enhance human interaction, not replace it. The tools should serve the care, not the other way around.
AI demos can look incredible. They can summarize documents in seconds. They can generate workflows. They can answer questions. They can make a prototype look polished and production-ready. But Jim cautioned carriers not to stop at the demo.
He recommended peeling back the onion and looking deeper, asking: What happens behind the scenes?
That means carriers need to understand how the AI is built, what data it uses, how it is prompted, how it performs under real volume, what happens when it fails, and whether the vendor truly understands the claims environment.
A prototype may look impressive. But claims is not a low-stakes environment where a flashy front end is enough. Jim was direct about this point: in a regulated industry, “prototype” can be a concerning word if it is not backed by the right architecture, controls, and domain knowledge.
Carriers evaluating AI vendors should ask whether the vendor understands both sides of the equation: technology and claims. A vendor may understand AI workflows, but if they do not understand the carrier’s business, regulatory environment, claimant experience, and examiner needs, the solution may fall short when it moves from demo to implementation.
Another area carriers need to understand is cost. AI usage is often tied to tokens, credits, model calls, prompts, context, or volume. That means the way a system is designed can have a major impact on how expensive it becomes to operate. Jim gave a simple example: if a system sends 30 documents and a claimant’s full background into an AI model just to answer one narrow question, the carrier may be spending far more than necessary.
The question is not only, “How much does the AI tool cost?” The better questions are:
For non-technical teams, reducing AI cost may sound like simply “using AI less.” But Jim explained that cost control often comes down to using AI better. That includes better prompts, better architecture, better data selection, and a clearer understanding of the desired output. If carriers do not think about those things up front, they may later need to rework the system to reduce cost, which can create operational disruption.
Security was another major theme. Claims data can include sensitive personal, medical, financial, and identifying information. When AI is introduced, carriers need to know exactly where that information goes and how it is used. Jim cautioned that if a carrier does not know how its data is being used, it has to assume risk.
Some AI tools may use submitted data to train models. Some providers offer agreements or configurations designed to prevent customer data from being used for training. Some cloud providers and AI companies offer enterprise-level protections. But carriers cannot assume those protections exist automatically.
They need to ask.
That is especially important when AI is used for document summarization. Claim forms and supporting documents may include PHI, PII, Social Security numbers, medical history, beneficiary information, addresses, and more. Carriers should understand:
Jim also noted that future regulation around AI is likely. Carriers should not wait until regulation arrives to begin thinking about responsible use. They should be asking now how vendors are preparing for that future.
Jim’s closing advice for carriers was to be cautiously optimistic. AI is not something to ignore. It is here, and teams will need to learn how to use it well. Carriers should not chase AI simply because it is shiny. They should not assume AI can do everything. And they should not overlook the complexity of implementing AI safely and effectively in a claims environment.
Instead, they should start with their process. Find the ineffective points. Identify the work that slows teams down. Determine whether technology can help. Then determine whether AI can make that technology stronger. That path is more practical, more sustainable, and more aligned with the real purpose of claims transformation.
AI has tremendous potential in claims. It can reduce manual work, improve speed, support examiners, and help carriers better serve claimants. But the future of claims should not be AI replacing people. It should be AI helping people do their best work.
If your team is exploring AI in claims and trying to separate practical opportunity from vendor noise, Benekiva would be happy to be a sounding board.
Whether you are evaluating vendors, reviewing internal use cases, or simply trying to ask better questions, we would love to help you think through how to approach AI wisely, safely, and effectively.
