The insurance industry faces a pivotal moment. As agentic AI - autonomous systems capable of reasoning and making real-time decisions with minimal human intervention - matures beyond prototype stage, major global insurers must navigate a complex landscape of opportunity and risk. How quickly they adopt this technology, where they deploy it and how they manage the organisational implications could reshape insurance distribution, underwriting and customer experience for decades to come. The conversation around agentic AI adoption is accelerating rapidly, with some insurers already moving from testing to real-world deployment.
The notion of autonomous agents even engaging in live conversations with customers is rapidly moving from theoretical to practical. Insurtechs are leading the way, testing agentic AI agents capable of sophisticated customer interactions, including live conversations.
Traditional insurers are watching closely and preparing to follow. The competitive pressure is evident: global insurers who don't learn from these early implementations risk falling behind players more accustomed to rapid iteration. Organisations across the industry are beginning to evaluate where and how agentic agents might transform their own customer interactions and internal workflows.
Some of the world's largest insurers are already using some forms of agentic AI. This month, Allianz officially launched its first agentic AI system: Project Nemo, in Australia. This AI application is designed specifically to streamline food spoilage claims. The insurer is now planning “a global ecosystem of specialised AI agents working alongside human experts.
This deployment signals a broader and global industry shift. According to some recent analysis, insurers who successfully scale agentic AI for dynamic workflows can realise productivity gains of 20-30 per cent (20-30%). The economic case is compelling.
Tim Kane, head of retail at Zurich Australia & New Zealand, leads strategy around agentic AI. In a recent panel discussion with insurance technology leaders, hosted by Kanopi and Endava, he said that insurers must act now based on market dynamics.he said that insurers must act now based on market dynamics.
"We're looking at agentic AI systems, where they should be sitting in our tech stack, what it means for our distribution model," he said.
The technology isn't just an incremental improvement to existing processes - it's a fundamental rethinking of how insurance reaches customers and how internal teams work. The competitive pressure is real. Insurers, who don't adapt, risk losing distribution flexibility to nimbler competitors, whether traditional players or new entrants.
The question of where agentic AI should operate in an insurer's technology infrastructure matters enormously. Early implementations tend to sit on the edge - handling customer interactions, automating routine tasks, accelerating well-defined workflows. But Kane suggests the endgame is more sophisticated.
"At this stage where agentic AI systems have been implemented, they're more likely to be sitting on the edge," he said. "[But] where we want to move to, is it becoming more of the core."
A hybrid approach, where agents both interact with customers and handle complex decision-making tasks orchestrating data across systems, represents the direction Zurich and likely other sophisticated insurers are heading.
This architectural evolution matters because it signals that agentic AI isn't just a customer-facing chatbot upgrade - it's becoming infrastructure that touches core business logic.
Agentic commerce - the use of autonomous agents in commercial interactions - opens possibilities that traditional insurance distribution has never achieved. Kane noted that "agentic commerce opens the door to new distribution models within insurance in general," with the potential for real-time insurance provision when customers actually need coverage. A traveller buying car insurance as they rent a vehicle, or a small business purchasing liability coverage the moment they sign a contract - these scenarios move from futuristic to feasible.
But the technology shouldn't be seen only as disruptive. Kane emphasised that agents can also make existing distribution channels, for example through brokers, more efficient. The transformation isn't necessarily about cutting out middlemen; it's about enabling all parties to operate at higher velocity and with better information.
There are a range of potential applications. Agentic AI systems are showing measurable impact across claims adjudication - where agents can manage entire workflows rather than simply scoring claims. Also in underwriting, where agents can evaluate medical questionnaires, health history and risk assessment. According to a McKinsey report, claims are showing a three-five per cent (3-5%) accuracy improvement, while new-agent success rates and sales conversion rates are improving by 10-20 per cent (10-20%) in some cases.
Technology adoption always succeeds or fails based on people. Kane stressed that "the people strategy is equally as important as any of the other strategies that we're looking at." The rollout mirrors how Zurich approached previous AI tools like Copilot: start with individuals experimenting and learning, then scale to team-level applications. For agentic AI, this means training employees not just to use agents, but to train, manage, and orchestrate them. The workforce isn't being eliminated; instead, workers are moving up the value chain from executing repetitive tasks to overseeing intelligent automation.
When it comes to underwriting and other judgment-heavy processes, Kane advocates for strategic human-in-the-loop decisions. Agents should handle certain journeys entirely, while humans should retain authority in processes requiring empathy, personal attention, or nuanced judgment. The key is choosing deliberately where automation excels and where human touch remains essential.
A crucial insight often overlooked in AI hype: deploying agents isn't the finish line. Then the labour shifts to agent management and orchestration. "There is a lot of work to make these agents effective," Kane emphasised. "You have to continually train them, work with them and provide continual feedback to make sure that you improve them day by day.”
For insurers just beginning to explore agentic AI, Kane recommended three critical questions:
1. Where can agentic AI enhance customer and employee experiences without losing the human touch that defines your brand?
2. How do you ensure trust, transparency, and compliance as these systems take on greater responsibility?
3. Are your data, technology, and people prepared to collaborate with agentic AI?
Kane predicted significant changes in insurance operations, particularly in acquisition and decisioning processes that could become "non-invasive, automated, real-time." But he urged caution: insurers must remain conscious of impacts on traditional business models and workforce implications, leaning into change management, rather than ignoring it.
