AI is reshaping long-held attitudes to build-versus-buy in digital insurance. The result is likely to be a shift away from one-time sourcing decisions towards the ongoing management of a technology portfolio that runs to a more flexible schedule than that allowed by traditional multi-year contracts.
In the past, the logic was often to build only systems that differentiated a firm in the marketplace, while buying in systems that supported commoditized functions. In this new world, a system that does not help to differentiate the firm may still be worth building, if agentic development tools sufficiently lower the cost of creating and maintaining the software.
At the same time, firms are questioning if all the processes locked inside traditional back-office platforms really belong there. Take claims, underwriting components, onboarding, and servicing, for example. These are the very foundations of the insurer's brand and the trust it builds with customers. AI makes it more affordable to bring these practices in-house where the insurer can give them the personal touch that customers value.
The challenges ahead
As AI reshapes the
- Economics: What the system genuinely costs to own over its lifetime
- Client trust: What is at stake in the interaction
- Brand: What might you reinforce, or erode through your decision
- Differentiation: To what extent can you stand out in a competitive landscape
For example, some firms are now trying out AI systems that handle insurance claims end-to-end. These fully automated agents have realistic voices and demonstrate a degree of simulated empathy. They are also, in theory, less expensive than human operators.
However, making the call to register an insurance claim can sometimes be a highly emotional experience. At a vulnerable point in their lives, customers are apt to notice the uncanny valley between an AI voice and an agent. For these claims, the answer is often to keep a person in the conversation, with AI doing the heavy lifting behind the scenes.
The cost you cannot see
AI-coded prototypes can be much cheaper to build, and prices are likely to keep falling. But a complex enterprise system in production costs more than the code.
Apply AI only to writing and testing code, and you ship fast into a pipeline that is slow upstream and downstream. The real advantage materializes when AI runs across the whole estate: data quality, security and performance testing, controls, production support, vulnerability management and resilience.
Swapping vendor lock-in for dependence on a single foundation model is another risk. Expect more model diversity and greater expectations for cloud providers. Self-hosted and air-gapped AI models that help to protect insurers from geopolitics or vendor commercial failure will likely grow in popularity – perhaps including smaller models that can run on-premises.
Cost discipline matters too. Frontier models are getting better, but they carry a significant price tag. Some big-name firms experience a situation where AI usage outpaces expectations. Not everything needs to run in real time on the most expensive model and insurers will increasingly pay close attention to the trade-off between speed and cost.
From forks to dials
Of course, this isn't the first time technology has unsettled the procurement process. Both the internet and the cloud disrupted the model, if only temporarily. But each of those waves only changed the inputs: what was newly possible to build, and what then became economic to buy. You still picked a side, signed for years, and lived with it.
AI changes the nature of the decision, not just its inputs. When building is fast, cheap and continuous, and the models and platforms beneath you keep shifting, insurers face a much more fluid situation.
If build versus buy was a fork in the road, AI has turned it into a dial. The firms that thrive will be the ones that know when to switch this dial up or down, decide each component on its merits and revisit that call as often as the technology demands.
The direction of travel is clear. Teams should revisit sourcing decisions continuously, rather than treat them as fixed multi-year commitments.
