The squeeze in underwriting profits is sustained across the insurance market, driven by persistently high loss and expense ratios. A combination of rising catastrophe activity, social inflation and increasingly complex commercial risks is compounding margin volatility. While the sector has returned to around sub-95 combined ratios, profitability remains highly sensitive to small shifts in the operating environment. This volatility, combined with growth opportunities including cyber risks, climate events, and liability coverage, has made underwriting discipline a necessity rather than a luxury.
This is where AI has become a focal point in the conversation, as insurers are looking to automate more of what matters and embed AI agents into business-critical processes to drive efficiency and protect margins. In theory, AI should help insurers process information faster, detect risk more accurately, and reduce leakage across underwriting and claims. In practice, however, the ambition to extend AI into complex knowledge work that traditionally required human judgment is not yet being realized at scale.
According to Camunda's State of Agentic Orchestration and Automation 2026 report, almost two-thirds (65%) of insurers admit there is a gap between their agentic AI vision and the current reality. While many firms report experimenting with AI agents, only 11% of projects reached production last year. If this pattern continues, insurance firms risk hitting an automation ceiling — one where AI agents compound operational complexity and fragmented IT systems, rather than strengthening underwriting profitability and reducing loss ratios.
Why AI still isn't trusted in insurance
Despite growing interest in how AI can support loss management and improve portfolio steering, trust remains a key barrier to scaling adoption in insurance. This is hardly surprising given the sector's exposure to regulatory scrutiny around customer protection, data security and operational resilience. At its core, insurance is built on accountability and risk transparency — every decision must be traceable, explainable, and auditable.
The majority of insurers (81%) are worried about the business risk of AI systems in day-to-day operations when IT teams lack adequate controls. A further 80% are concerned about a lack of transparency around how AI is used within business processes, while 68% cite compliance concerns and 63% lack internal skills to effectively manage AI.
AI adoption stuck in low-risk mode
While caution is natural in a highly regulated industry like insurance, it significantly affects where AI is deployed. Most agents remain confined to low-risk, isolated use cases, with 81% of insurers saying current deployments focus on chatbots or assistants that summarize or answer questions. These applications may improve efficiency, but they do not improve underwriting profitability or reduce claims leakage.
The challenge is that as AI agents move toward areas that directly influence margin stability, such as underwriting decision-making and claims adjudication, adoption becomes more constrained. These are key financial control points for insurers, where even small errors can lead to significant loss leakage or pricing inaccuracies.
However, if AI remains confined to the periphery of the business, insurance firms will fail to maximize returns on their AI investments. To break through the automation ceiling, insurers need a way to safely embed agents into margin-critical processes and ensure agents operate consistently within clearly defined parameters.
Bringing order to AI in insurance
Closing the gap between AI ambition and reality requires control over how AI behaves inside regulated insurance processes. The majority (90%) of insurers agree that AI must be orchestrated across business processes to get the maximum benefit from AI investments and ensure regulatory compliance.
Insurance operations already rely on structured, deterministic processes to manage underwriting decisions and claims routing, helping ensure accountability and traceability in high-risk financial decisions. The next step is extending that same level of discipline to AI, and this is where agentic orchestration comes in.
Rather than treating AI as a black-box tool, agentic orchestration combines deterministic guardrails with dynamic reasoning. This approach allows AI to participate in underwriting, claims, and servicing workflows while remaining within clearly defined guardrails that preserve auditability.
In practice, this creates a controlled environment where AI can adapt to new information, such as changes in claims severity or risk exposure, without compromising governance. Decision-making remains traceable, controlled, and aligned with regulation, while benefiting from the speed and accuracy of AI.
Scaling AI where it matters most
With combined ratios under pressure and loss volatility increasing, the insurance sector is looking to automation and AI to strengthen underwriting discipline and reduce P&L leakage without sacrificing transparency. There is also growing emphasis on activating a broader vendor partnership ecosystem, such as through insurtech alliances. As a result, insurers are looking to evolve their operating model, which will require a transposable orchestration platform that integrates, coordinates and scales insurance products across the value chain.
Agentic orchestration offers a path to move AI into the core of underwriting, claims, and portfolio management in a controlled and measurable way. This is where AI stands to have the greatest financial impact by improving pricing accuracy, reducing overpayment and strengthening consistency in decision-making.
Those organizations that succeed will not just be more efficient — they will be better positioned to protect margins, stabilize performance and maintain underwriting discipline in an increasingly volatile insurance landscape.
