Transforming Insurance with Agentic AI
Differentiating Agentic from Traditional AI
Basak Khalita, also known as BK, serves as the global head of financial services and insurance at Quantify, an AI-first digital engineering company. The discussion centers on how Agentic AI is transforming the insurance landscape from underwriting and claims to risk assessment and customer engagement, with insights on how insurers can adopt AI responsibly and drive measurable business value.
AI has been around in various forms for years. But Agentic AI in particular is being talked about as the next big leap. What exactly differentiates agentic AI from traditional AI approaches. Especially in the context of insurance, and why should insurers be paying attention today rather than waiting?
The Evolution of Business Technology
From Paper-Based to Digital Systems
To understand agentic AI in the context of evolving business needs and technology progress, consider the historical progression. Thirty to forty years ago, businesses moved from paper-based work. Where everything was written manually to applications that could store and process data electronically. This transition to electronic systems was primarily doing the same work that manual effort would do. But with 10 to 30% efficiency gains. Then came the digital world with more advanced tools like portals, web 2.0, and the discussion of 3.0. There was a significant transition into modes of communication using these technology platforms, with apps and other innovations emerging.
Radically Changing Engagement
What agentic AI will do and how it will differentiate is the fact that it will radically change the way people engage. In insurance, for example, if you have to get a risk binded and you want to get a quote out, there is a standard industry norm in practice where usually if it’s commercial insurance, the person goes to their agent. The agent will advise, visit the place, go to the carriers, fill up the form, and they get a quote in 3 to 4 weeks. That’s the usual standard norm. But agentic AI is going to do a variety of things in an asynchronous way in this whole value chain up front.
Autonomy and Intelligent Planning
Defining Levels of Autonomy
The business model will change. Agentic AI will transform the business. Technology will decide what kind of business flows will be in existence rather than business flows defining what kind of application needs to be built. It’s going to completely flip. It’s not just about intelligence or automation, but about AI making decisions and driving workflows autonomously.
Regarding levels of autonomy, from simple task automation to fully autonomous agents, insurers who are just beginning their agentic AI journey need to decide which level to adopt first and how to think about scaling over time. There are single task agents that perform a particular task every time they’re called upon. These are typically any kind of functions that anyone used to write programmatically or similar to an RPA bot that knows a set of things to be done but will only do those set of things.
Agent Communication Protocols
Agents will typically be intelligent enough to know, decide, and plan their tasks. That will be the biggest difference that agents will bring. This is the level of autonomy that leads into agents talking to agents. There are now many MCP gateways and protocols in place. Agents talking to agents will be the most critical aspect. There are four different categories of agents, and agents will decide the way they want to interact at what levels.
Starting with smaller contained workflows helps build confidence and demonstrate value to stakeholders before moving to larger scale initiatives.
Success Stories in Insurance
Automating Claims and Quotes
Looking at the whole insurance value chain, there are three primary journeys: a quote journey, a policy journey covering all life events of an insured, and finally, if there is a claim, how to manage that journey. In each of these journeys, claims followed by quotes have been the two journeys where agentic is making an initial impact, with involvement in numerous such implementations.
This includes creating autonomous claims processes. Organizations can pick and choose those parts of a claims process which can be automated if they don’t have the organizational buy-in to do a wholesome change but start small. There have been measurable successes in making sure FNS are being done appropriately and coverage validations are being done. There are numerous agents that have shown success.
Optimizing Submissions and Policies
On the quoting side, taking care of submissions, especially for those lines where it takes considerable time and effort, the submission process is being completely automated, leading to getting the data in the hands of the underwriter at a much faster pace. There have been real successes in these two primary areas. Policy, given the fact that the journeys are not that standard and depend on how the insured moves, hasn’t seen too many use cases popping up, but that will be the third frontier.
Responsible AI Governance
Industry-Specific Evaluation Dimensions
Insurance is highly regulated, so responsible deployment is critical. How can insurers implement agentic AI while maintaining compliance, governance, and trust without slowing down innovation? That’s where significant effort will go and needs to go. When discussing business processes or applications that can decide on their own what to do, unless there are checks and balances in place, it will never be possible to get past the commissioner.
What needs to happen is that evaluation will have to get contextualized from an industry standpoint. Looking at evals from an agentic AI standpoint, there are three to four different dimensions. One is checking on transparency and explainability, with various scores around it for evals.
Safety and Policy Compliance Pillars
Then there’s fairness and policy compliance. Is it complying with the organization’s policy? Is it able to resist any malicious prompts being thrown at it and not reacting? And finally, the safety pillar where failure containment scores need to be examined. These evals are important, but beyond that, what is critical is bringing the business context. If putting an eval for an AI agent application on the quoting side, there’s an absolute need to focus on fairness and policy compliance because there are things that can go wrong.
The Hub and Spoke Model for Scale
Governance doesn’t have to slow things down. It’s about putting the right framework in place. There are different models in place. The ones primarily being adopted by customers are the hub and spoke model, but in a different way. The hub and spoke model used to be primarily with the hub being more governance, directional view, architectures, and the spokes are where actual real work happens. But in an agentic ecosystem, the hubs are becoming the providers of the technology landscape. When trying to build an agent ecosystem, before going to those agents, the foundation is needed: repeatable templates, inference engines, and other components need to be in place to build agents at scale. The hubs in many customer organizations are becoming the provider of that particular platform or the tech which the spokes can use to build their solutions.
Adoption Models and Human Factors
Overcoming Human Skepticism
Technology is only half the story. Another roadblock for companies is often adoption, and adoption depends on people. What strategies ensure that underwriters, claims adjusters, and other teams actually embrace these tools?
The level of skepticism whenever bringing a new technology always exists. It is nothing new. It happened in the past as well. There will always be people being skeptical of what value it can deliver. Agentic AI is no different. At the same time, it also brings hope that it can do things that are not known right now.
AI as a Collaborative Assistant
From an adoption standpoint, clients will be looking out for early successes. Making sure the right success criteria is established and the hypothesis being tested is well explained so that all these key stakeholders or personas can adopt it. They should see this more as an assistant’s ability to help them. The analogy of runners who are now using AI-enabled plates that help them run faster applies here. Think of agentic AI as something that can only help perform tasks better and faster. People are more likely to use technology that actually makes their work and their life more efficient.
The Future Landscape of Insurance
New Business Operating Models
Over the next few years, how will Agentic AI actually reshape the insurance landscape, particularly in underwriting, claims, and customer engagement?
Gone are the days where business needs will drive the technology. The available technology and its possibilities will create new business operating models. That’s the difference with agentic AI. The ability and the scale at which agents will perform will drive business leaders to think about newer workflows, optimized processes, ability to get customers faster, serve them better, and be nimble in product selection. There might be entirely new business design processes being put in place through agentic AI.
Identifying Future Risks
Massive changes are anticipated. Insurance was not something that a customer used to interact with every day. It’s more of a once in 6 months or once in 12 months interaction. But with agentic AI coming into the picture, it will become much easier to do various tasks. The carrier community needs to do significant work in terms of identifying the risks of the future. Looking at the products that have been out there for the last 50 to 70 years, there might be newer products popping up, and that’s where agentic AI will push the industry toward.
Transformation and Closing
It’s exciting to get a glimpse of a future where AI isn’t just a tool but a core part of how insurers operate and compete. It’s clear that the right strategy, governance, and human-centric approach will lead to insurers being able to transform their operations and unlock significant business value through AI. The technology is expected to continue to evolve in the industry, with BK from Quantify providing valuable insights into this transformation.
