As the insurance industry grapples with the rapid evolution of artificial intelligence, the long-standing debate between building proprietary software and purchasing off-the-shelf solutions is undergoing a radical transformation. Simon Glairy, a distinguished authority in Insurtech and risk management, brings his extensive experience in AI-driven risk assessment to explore how this technological shift is redefining the operational foundations of financial services. This discussion explores the transition from static, multi-year procurement cycles to a more fluid, portfolio-based approach to technology management. Key themes include the reduction of development costs through agentic tools, the strategic repatriation of customer-facing functions like claims and underwriting, and the complex balance between automated efficiency and the essential human touch in high-stakes interactions.
How do agentic development tools fundamentally alter the traditional logic of building only what differentiates a firm in the market?
In the past, the strategic math was relatively straightforward: an insurance firm would build only the systems that provided a unique competitive advantage while buying commercial solutions for commoditized back-office functions. Today, agentic development tools are slashing the overhead of software creation and maintenance so significantly that this “buy” default is being aggressively challenged. Even a system that doesn’t offer a unique market edge may now be worth building in-house if the cost floor has dropped low enough to make custom development economically viable. This allows insurers to move away from being held hostage by traditional multi-year contracts, favoring a technology portfolio that runs on a much more flexible and responsive schedule. By owning more of their stack, firms can ensure their tools evolve exactly when the business needs them to, rather than waiting for a vendor’s roadmap to align with their goals.
Why are we seeing a renewed push among insurers to bring foundational processes like claims, onboarding, and underwriting back in-house?
There is a growing realization that processes like claims, underwriting, and customer servicing are not just administrative tasks; they are the very foundations of an insurer’s brand and the bedrock of the trust they build with their clients. AI makes it far more affordable to bring these critical practices back into the internal fold, where the insurer can provide the specific personal touch and brand consistency that customers value. When these functions are locked inside a third-party platform, the insurer loses the ability to differentiate the customer experience at the moments that matter most. By leveraging AI to handle the operational weight of these departments, firms can reclaim control over their brand identity and ensure that every interaction reinforces their unique market position. This shift is about moving away from generic, one-size-fits-all servicing toward a model where every touchpoint is a deliberate reflection of the company’s values.
In the context of automated claims agents, how do you balance the cost-efficiency of AI with the need for human empathy during a customer’s time of crisis?
We are currently seeing firms experiment with fully automated agents that use realistic voices and demonstrate a surprising degree of simulated empathy, which are, in theory, far less expensive than maintaining large human operator teams. However, the act of filing an insurance claim is often a raw and highly emotional experience for a customer who may be at a very vulnerable point in their life. There is a significant risk of falling into the “uncanny valley,” where a customer senses the artificial nature of the AI voice and feels alienated or undervalued by the machine-led interaction. The most effective strategy we are seeing involves using AI to do the heavy lifting behind the scenes—processing data and checking policy details—while keeping a human being as the primary face of the conversation. This ensures that the customer receives the genuine empathy they need while the system maintains a high level of technical efficiency.
Beyond just writing code, what are the hidden complexities and costs of moving an AI-driven system from a prototype into a full-scale production environment?
While it is true that AI-coded prototypes can be built for a fraction of the traditional cost, a complex enterprise system in a live production environment carries costs that go far beyond the initial lines of code. If a firm only applies AI to the writing and testing phases, they often find themselves shipping software into a pipeline that remains frustratingly slow both upstream and downstream. The real competitive advantage only materializes when AI is integrated across the entire estate, covering data quality, security testing, vulnerability management, and production support. Insurers must also exercise strict cost discipline, as the usage of high-end frontier models can quickly outpace expectations and lead to unexpected budget strains. It is no longer just about the cost of development, but the total lifetime cost of ownership, which includes the continuous trade-off between model speed and the price of compute.
What emerging risks should security teams prioritize as they integrate more diverse AI models and deal with the rise of “Shadow AI” within their organizations?
The landscape of risk is expanding because every new agent, model, and privately-introduced tool represents a potential new entry point for cyber incidents. We are seeing a move away from simple vendor lock-in toward a potentially more dangerous dependence on a single foundation model, which is why model diversity and the use of air-gapped, self-hosted models are becoming so popular. “Shadow AI,” where employees use their own unauthorized models for work tasks, brings severe risks of data exfiltration and prompt-injection that most firms are not yet equipped to handle. There is a desperate need for mature, enterprise-wide visibility products that can monitor how AI is being used across the entire firm in real-time. Currently, both in-house security teams and external vendors are in a high-stakes race to develop the tools necessary to provide this level of oversight and protection.
How does the transition from viewing sourcing as a “fork in the road” to viewing it as a “dial” change the way technology leaders manage their long-term strategies?
Historically, technological shifts like the internet or the cloud changed the inputs of the decision—what was possible to build and what was economic to buy—but you still ultimately picked a side and lived with it for years. AI is different because it changes the very nature of the decision-making process by making building so fast and cheap that the ground is constantly shifting beneath your feet. If the choice between building and buying was once a fork in the road, it has now become a dial that a firm must be prepared to turn up or down as the technology and market conditions evolve. The insurance companies that thrive in this new era will be the ones that treat sourcing decisions as fluid and continuous rather than fixed, multi-year commitments. This requires a cultural shift toward constant reassessment, where every component of the technology stack is regularly evaluated on its own merits in light of the latest AI advancements.
What is your forecast for the future of AI-driven risk management in the insurance sector?
I anticipate a significant move toward the use of smaller, specialized AI models that can run on-premises or in air-gapped environments to protect against geopolitical instability and vendor failure. The industry will move away from a “bigger is better” mentality regarding models, focusing instead on the trade-off between the high cost of frontier models and the need for real-time, efficient performance. We will also see cyber security and AI management become completely intertwined, as the “Shadow AI” problem forces firms to implement much more rigorous, enterprise-wide visibility tools. Ultimately, the successful insurer of the next decade will not be the one with the most advanced AI, but the one with the most disciplined and flexible approach to managing their technology portfolio. Those who can master the “dial” of building and buying will be the ones who maintain their edge in an increasingly automated and competitive landscape.
