As artificial intelligence continues its relentless march towards matching various facets of human intelligence, the insurance industry must pay attention to an emerging capability that could transform application development. Within the next 3-5 years, AI systems may be able to watch video captures of application user flows and automatically generate functionally equivalent code.
This could immensely accelerate insurance IT projects and lower application development costs. However, while the underlying deep learning advances make this feasible soon, businesses may need to temper expectations around maturity and manage risks.
Key Advances Building Towards This Capability
In the last few years, neural networks have achieved impressive results in analyzing images and video, understanding natural language, logical reasoning, and generating code. Combining these strengths, AI agents could potentially map UI flows and data logic captured from application demos onto the underlying codebase.
Vision systems can now identify on-screen components, actions, and responses. Language models allow translating text and speech into structured knowledge. Reasoning algorithms can map between conceptual representations. And code generation techniques have progressed from simple programs to complex architectures.
Assembling these capabilities into an integrated solution is the next step several AI teams are eyeing. The payoff for automated coding at scale is substantial, evidenced by heightened VC interest in startups like Anthropic working on this technology.
Emergence of New Competition
However, the same capability allowing rapid application development from demonstrative videos could significantly disrupt established technology vendors serving the insurance industry. As barriers to entry into solution creation lower, new players leveraging AI coding could launch competitive alternatives faster and cheaper.
Incumbents may suddenly find their extensive software IP and armies of developers losing differential value. Industry expertise encoded into proprietary systems over years risks being commoditized by AI-fueled startups. Emerging “no-code” tools empowering non-programmers present an existential crisis for current market leaders.
Mitigating Strategies for Protection
Legacy technology companies can adopt defensive innovation strategies to protect market position. Investing in internal AI coding labs could help to compete with external offerings. Acquiring promising startups early in capability development further aids maintaining competitive parity.
Strategic partnerships with AI platform companies, academic labs, and integrators also help monitor, evaluate and ultimately deploy automated coding alongside human developers. Proactively evangelizing tools, use cases and integration blueprints engenders ecosystem support.
Carefully navigated, AI-based software automation could expand markets for new solutions while responsibly balancing job impacts across the insurance technology value chain.
The Path Ahead
Neither hype nor despair is currently warranted when evaluating automatic coding from app demos. What insurance technologists should track is how incremental breakthroughs in causal visual reasoning and software synthesis stack up year-on-year.
When machine learning models reliably demonstrate the ability to understand complex logic from limited samples, such as a video of an application screen flow, the software development process will be transformed providing an order of magnitude improvement to software development productivity. Continued research paper releases and technology validation experiments will illuminate how close this future actually is.
More advances are required before the core systems vendor community is disrupted and the best option for most insurers remains a vended solution. If you’d like to discuss your policy admin, claims or billing core system needs, please contact me or Mitch Wein to continue the conversation.
