TechTalk: Agentic AI presents a sink or swim moment for insurance sector

Artificial intelligence (AI) remains a hype generating buzzword, but has become so commonplace in business strategies that it is almost omnipresent. Since the end of 2022, when the consumer version of OpenAI’s ChatGPT model was released to the public, the technology world has been awash with this type of generative AI and the capabilities that it provides. 

And while this technology has become something many in the insurance market are looking to leverage across numerous business use cases, there is a new kid on the block that is making a name for itself. 

Agentic AI has begun to garner considerable interest from the business world for its potential to create synthetic workers and supplement human teams with digital expertise.

In January 2025, Jensen Huang, chief executive at multitrillion market capitalisation technology firm Nvidia, told delegates at technology conference CES 2025 that the move from generative AI towards “the rise of agentic AI” was the world’s next major technological shift and would create a multitrillion dollar opportunity for businesses across the globe. 

Definitionally, agentic AI is differentiated from the now well known generative AI – which is used to create new content – by its ability to autonomously set, pursue and complete tasks and make decisions independently.

A World Economic Forum white paper from December 2024, entitled Navigating the AI Frontier: A primer on the evolution and impact of AI agents, explained: ”An AI agent responds autonomously to inputs and its reading of its environment to make complex decisions and change the environment without constant human intervention and authority.”

As a business type, insurance is fertile ground for the benefits that this nascent form of AI can provide, with agentic AI’s ability to ingest information and quickly process it allowing for efficiencies in areas like claims handling, complaints management, customer service and even underwriting. 

However, it is exactly this suitability to agentic AI that makes it paramount for insurance sector firms to be fully ready to take advantage of the technology. 

Speaking at March 2025’s Insurtech Insights conference, Simon Torrance, founder and chief executive at advisory firm AI Risk, explained: ”The biggest risk for insurance companies, and indeed for any company, is being outcompeted by your closest rivals that can exploit agentic AI better than you.

”In five years’ time, for example, a mid-sized insurance company with 5,000 human workers could add 10,000 digital synthetic workers to its workforce and essentially triple its operational capacity.

”People are the biggest asset in insurance and when you have teams of workers – in this case, digital workers – undertaking activity much quicker and cheaper than humans, in combination with a human workforce, then you can achieve wonderful things.” 

Fertile ground 

There is undeniable hype around agentic AI and, like with any new technology, cutting through this to identify real use cases can be difficult.

However, agentic AI is already in use across the sector, with insurance presenting perhaps one of the best crucibles in which to test the efficacy of the technology.

Speaking to Insurance Times, Quentin Colmant, chief executive and co-founder at insurtech Qover, noted: ”Insurance is one industry where the product is 100% defined in the terms and conditions and the policy schedule of the client, so it’s actually an amazing use case for agentic AI.”

Colmant’s firm is walking the talk here too, with Qover recently rolling out AI agents to some of its embedded insurance clients, such as Revolut and Deliveroo. He explained that, in some of Qover’s programmes, 20% of total customer queries were being fully answered and solved by agentic AI.

He added: ”It’s an amazing gain in terms of efficiency but, beyond that, the accuracy of the answer is actually on par or better than humans, following a close review.”

Andy Moss, chief executive and co-founder at commercial insurance software firm Send, agreed that insurance presented an excellent use case for agentic AI. 

He explained: ”The suitability of insurance to agentic AI is significant, especially for complex risks and commercial specialty business where there are generally lots of documents and emails flowing around.

”There’s a lot of manual handling and processing in insurance, as well as things you need to remember to do or not to do. Having a digital co-worker who’s keeping an eye on some of those things for you and actually doing some of the work presents an obvious use case and is now something that is in scope for the technology.” 

Moss continued that an AI agent could assist with the prioritisation of submissions from brokers much more effectively than current systems, for example, or even handle issues around letters of authority and the necessity for referrals. 

Colmant added that one use case Qover was currently utilising agentic AI for focused on complaints handling.

When a complaint is received about a customer’s insurance, Qover activates three AI agents to assist the complaints handler – an agent that acts as a lawyer for the customer, another that acts as counsel for the insurer and a final agent that judges the arguments of the other two. 

The discussion and arguments that each AI agent makes, as well as the judgment the judge agent comes to on those arguments, are then made available to the human agent, who can use that information to come to an informed decision on the legitimacy of the complaint, based on a full understanding of the customer’s issue. 

Colmant added that this process could also easily be applied to claims management, with this method having already reduced the cost of claims for some of its clients by half.

He explained: ”At a standard insurance company, around 60% of the staff are operational people doing customer care and claims management, so this is not only about efficiency and cost reduction, but about the quality of servicing and speed of processing.”

All about implementation

For all of the advantages that agentic AI could bring to business processing in the insurance sector, it is vital to cut through the hype and focus on effective implementation. 

Moss explained: ”You can’t just start using this stuff. You’ve got to plug it in somewhere and have something to hang it off of to actually operationalise the technology – and I wouldn’t underestimate how hard that can be.” 

He added that while the technology now has powerful capabilities, actually bringing agentic AI into workflows at an insurance company would be a slow process of experimentation and implementation. 

“The tech is there, now it’s about the hard work of bringing it to the industry and being successful with use cases that matter to your business,” he said.

Adam Miller, chief information officer at Markerstudy, echoed this assessment. He noted that while it was incumbent on chief information and technology officers to be exploring agentic AI, it was “early days” and most firms were at the stage of identifying frameworks where it could be utilised effectively. 

Miller added that, despite the “hype cycle”, barriers to the implementation of the tech would also present themselves over time, especially issues such as consumer trust of artificial agents and the transparency of information used in decision-making processes required by regulation. Training people to work with synthetic colleagues would also require significant training, he said. 

Outside of the minutiae of adoption and implementation, Colmant said that the prompt design and training of AI agents was also essential to get absolutely correct.

He explained: ”The output of any AI agent will be highly dependent on the instructions you give them, so whenever we create a new one, we treat them as a new employee who knows nothing about the company or even insurance.

“You need to define a very strong prompt to teach them what they need to know, especially if they’re given a specific function. That doesn’t end once it’s launched either because you then have to work a lot in the weeks and months after release to always optimise the prompt.” 

There is no doubt that AI agents will soon proliferate across the insurance industry and bring significant improvements to many processes and consumer outcomes – but there is the hurdle of adoption and implementation to be surmounted. 

What is clear, however, is that insurance companies cannot allow this technology to come as a surprise – the advantages accrued to early, effective adopters will be crucial, meaning that ignoring agentic AI’s potential is a significant risk.