Credit: YouraPechkin/Adobe Stock
For years the insurance industry has repeated the same promise about AI: automate enough tasks and you reduce operating costs. But if you talk to people inside carriers, brokerages, and TPAs, a very different pattern is emerging.
Instead of lowering workforce expenses, AI is driving them up. And not because the technology is failing, but because the way we use it is out of sync with how insurance work actually gets done. A recent survey found that nearly 90% of insurance executives now identify AI as a top strategic priority, yet only about one in five have solutions running in production, which means most organizations are paying for experimentation without seeing scaled efficiency gains yet.
The myth is that automation replaces labor. The reality is that automation changes labor, usually faster than organizations are prepared for. And that gap between expectation and reality is exactly where insurers are seeing their biggest risks, operational bottlenecks, and unplanned costs.
Let’s start with the most common assumption: that AI will shrink claims teams. In practice, the opposite is happening. AI can summarize a claim file, pull loss information from unstructured documents, and even draft the first version of a communication to the insured. But every one of those AI-generated outputs requires a human to validate it.
Claims professionals tell me they are spending less time collecting information and more time auditing it. When AI is plugged into environments shaped by regulation, litigation, and strict documentation standards, nothing can be taken at face value. You speed up intake, but you add work to review, correction, and quality control. Teams are moving faster, but they are not becoming smaller.
The same pattern is playing out in underwriting. AI can prefill applications, flag missing data, and surface historical loss patterns. But the moment AI becomes part of a risk decision, the liability shifts to the underwriter. That is why underwriters are spending more time explaining decisions, documenting the model’s role, and double checking outputs against internal guidelines. AI does not remove underwriting complexity. It exposes it. Underwriters become editors of machine-generated work rather than authors of every step, and that shift increases responsibility, not reduces it.
Across the board, insurance teams are running into what I call the “second-order cost of automation.” The first-order cost is the investment in tools. The second-order cost is the work created by those tools. In insurance, the second order cost is larger.
Automation speeds up parts of the workflow, which means the rest of the workflow becomes the bottleneck. A claims file that used to take an hour to assemble now takes 10 minutes, which puts pressure on the adjuster’s review time. A submission that used to take days to process now hits an underwriter’s inbox fully organized within minutes, which shifts bottlenecks downstream to risk evaluation and documentation. Workers are not being replaced. They are being asked to operate at a higher cognitive level, with more accountability, in less time.
Here is the part the industry does not talk about enough: most insurers are trying to use AI without changing the underlying operating model. If you automate tasks but keep the same handoffs, documents, approval flows, and workload expectations, you get the worst of both worlds. You pay for automation, but you continue to work as if nothing has changed. That is where the real cost inflation comes from.
There is a better path. Instead of asking, “What tasks can AI automate,” leaders should be asking, “What decisions should humans still own?” That one question reframes how you deploy automation. When you identify where human judgment is essential, you also identify where automation can safely compress work without creating new risk.
Claims triage, low-severity FNOL processing, correspondence drafting, benefits eligibility checks, and internal research are examples where AI removes manual burden without introducing regulatory exposure. High-severity claims, coverage analysis, subrogation strategy, and risk selection need human ownership, with AI serving as an advisor rather than a decision-maker.
The takeaway is not that AI is failing. It is that the insurance workforce was not designed for a world where machines do the first draft of every task. Carriers and brokers who acknowledge this are already redesigning work around people, not around tools.
They are clarifying decision rights, separating human-required steps from machine-accelerated ones, and building governance frameworks that match the industry’s regulatory reality. They are discovering that the goal is not fewer people. The goal is people doing better, more strategic work.
AI will not reduce workforce costs unless insurers change how work is structured. But if they do, the payoff is not a headcount reduction. It is capacity. Faster cycle times. Better documentation. Fewer errors. And teams that can finally move beyond repetitive tasks toward the judgment-heavy work the industry depends on.
AI will not eliminate insurance professionals. But it will reward the organizations that redesign the job to match the tools, rather than expecting the tools to magically redesign the job.
Opinions shared in this piece are the author's own.
Frank Palermo is the Chief Operating Officer of NewRocket, where he helps guide the company’s growth strategy and strengthens its position as a leading advisor in digital workflows, AI, and enterprise transformation. He brings decades of experience building and scaling technology and consulting organizations, with a career that spans software engineering, enterprise platforms, cloud, data, and AI-driven services. Frank is known for combining deep technical fluency with clear operational vision, and for helping clients translate modern technologies into meaningful business outcomes.
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.