Credit: wenich/Adobe Stock
Most insurance organizations—carriers, brokers, wholesalers—are still sitting on the fence when it comes to AI.
It’s not an opinion, it’s a fact; McKinsey reports that the majority of insurance companies are still experimenting rather than deploying AI in production. McKinsey: The Future of AI in Insurance, and BCG studies found that only 7% of insurers have managed to scale AI across their organization.
For insurance professionals, to keep it simple: this is bad. Hesitation is becoming a competitive liability, and those who rely on processes that “work for now” are falling behind those who are learning and adapting.
So what’s the hold up? The reluctance stems from four concerns I hear across the industry: accuracy, liability, budget, and a fear of overpromise. Each is valid, but none can be solved by waiting. In reality, these concerns are only resolved through adoption.
Accuracy—you can’t optimize what you don’t deploy
Many insurance teams expect AI tools to work like plug-and-play software, but that isn’t how these systems mature, especially in regulated industries with complex workflows.
There is no perfect time to deploy, and accuracy only improves once organizations commit hours, resources, and people to training, using, and supervising the tools. That is how any tool or major enterprise, both inside and outside of insurance, achieves reliability. The only thing “waiting for the right moment” does is delay the timeline of dependable outputs.
Budget: AI is a labor cost, not a license cost
One challenge is that many teams evaluate AI pricing like traditional software procurement. But AI is a different category entirely, and here’s why:
● Its potential return on investment is significantly higher.
● It improves rapidly over time (compare today’s consumer AI tools to those available just six months ago).
● Many solutions use usage-based pricing, meaning cost adjusts with utilization and the value provided.
When you compare AI to the cost of hiring, rather than the cost of a software license, it suddenly looks inexpensive. Add to that the fact that many vendors offer month-to-month plans, so brokers and insurers can test things out without locking themselves into a long contract.
Overpromise vs. Underdeliver: Understanding AI deployment
Industry conversations are often centered around AI hype cycles. While overpromising is real, what many overlook is that early friction (such as natural hiccups around integrations, workflow redesign, and forward-deployed engineering) is actually part of creating long-term value.
McKinsey’s findings support this: insurers who move past experimentation and into real deployment gain measurable operational uplift. In my experience, these early challenges rarely indicate failure. Instead, they show that the organization has started the transformation process.
The future of human-AI hybrid teams
Capgemini Research Institute predicts that by 2028, 38% of organizations will have blended teams where human professionals work alongside AI agents on key workflows. For the insurance industry, this matters a great deal.
Collaboration between humans and AI will improve accuracy, speed up processing, and help address long-standing concerns around trust and liability. As AI continues to improve, costs are likely to drop and reliability will rise, but only for organizations that start building experience with these tools early.
If you’re sitting on the fence because your current process “does the job,” consider what your competitors are doing and what your organization could be missing. Every month is a month other companies spend training models, refining workflows, and optimizing their systems, widening the gap between those who adopt AI and those who don’t. Once you make the decision, you should prepare for a couple of months of evaluation, onboarding and internal adoption.
The insurance companies that come out ahead won’t be the ones waiting for perfect conditions—they’ll be the ones willing to learn and adapt. The next step is to move beyond exploration and start controlled deployment. Identify a workflow that has a high manual workload but low regulatory risk, run a pilot with real data, and dedicate resources to training the system.
AI is a tool, and the organizations that start using it effectively today are the ones that will shape the industry over the next decade.
Opinions shared in this piece are the author's own.
Aman Raghuvanshi is the Co-Founder and CEO of Pyq, a Y Combinator-backed company transforming commercial insurance operations through intelligent automation. aman@pyqai.com; (909) 529-4371
© 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.