Artificial intelligence (AI) is transforming software development into the key differentiator for commercial insurance, moving beyond just pricing. John Swigart, co-founder and CEO of Pie Insurance, warns against solely licensing third-party software solutions, arguing that AI cost-effectively democratizes the ability to build proprietary tools in-house. This capability, once limited to carriers with large engineering teams, now makes in-house software development crucial for competitive differentiation, as carriers relying only on vendor solutions risk falling behind and appearing identical to their competitors.
Why bespoke models still require scale
In the small commercial insurance sector, AI investment is intensifying the 'buy-versus-build' debate concerning modeling software. Swigart advocates for bespoke underwriting models, particularly where a carrier possesses genuine specialization within its book of business. However, developing such proprietary models necessitates access to massive volumes of data, often including external third-party data that must be licensed. This constitutes a significant barrier for smaller carriers, as building and maturing these models requires substantial upfront capital investment and the ability to absorb underwriting losses during their development. Furthermore, commercial lines are considerably less developed than personal lines, like auto, in terms of third-party data availability. The vast diversity of business operations, exposures, and classifications in commercial risk makes modeling more complex. This complexity is a primary reason why smaller commercial carriers frequently depend on outside vendors for data and technology, rather than developing their own in-house solutions, leading to surprisingly high levels of manual operational work even in technologically advanced firms. Achieving the necessary scale for proprietary models is challenging; for instance, Next Insurance, a rival small-commercial insurtech, took a decade to accumulate over 600,000 policyholders and significant revenue to build its extensive underwriting and loss data sets.
Where the productivity gains are real – and how to find them
Despite the structural constraints, Swigart is confident about the significant productivity gains that AI adoption can bring to commercial lines, especially through in-house solution building. He highlights 'real value' in enhancing individual and company-wide productivity, empowering each role to achieve greater leverage. Swigart points to improvements across analytical and operational tasks, with increasing potential in agentic AI use cases as the technology matures. His key advice for leaders considering AI implementation is to avoid excessive planning and instead focus on starting, experimenting, and finding practical use cases. This approach aligns with Deloitte-cited research, which indicates that early agentic AI deployments in insurance have yielded impressive results, including underwriting efficiency gains of up to 36% and claims cycle-time reductions nearing 40%.
The data problem specific to small commercial
Swigart identifies firmographic data—accurate and continuously updated information about insured businesses—as the clearest emerging solution opportunity. Different commercial coverage lines require specific segments of this data for effective pricing and underwriting. For example, workers' compensation necessitates precise classification of business activities (e.g., distinguishing a commercial plumber from a residential one), as misclassification directly impacts pricing and loss experience. The National Council on Compensation Insurance maintains roughly 700 distinct classification codes, and a single misclassification can alter workers' comp premiums by thousands of dollars annually. Swigart suggests that advances in AI technology can enable carriers to analyze streaming firmographic data to identify crucial business signals and risks in real-time, preventing issues from being discovered only at renewal or during a claim. Ultimately, AI is compelling small commercial insurers to focus their in-house investments on four key areas for competitive differentiation: developing proprietary software instead of wholesale licensing, creating bespoke underwriting models supported by licensed external data where scale permits, adopting a broad operational AI strategy driven by experimentation, and establishing precise firmographic data pipelines to proactively identify classification errors. Swigart concludes that relying solely on third-party software will result in carriers looking identical to competitors, making in-house building essential for establishing a unique market edge.
