Shepherd’s $42M Series B might look like another insurtech funding announcement at first glance. The more interesting angle sits beneath the headline. The company is not trying to broadly improve insurance. It is focused on a very specific bottleneck: underwriting for large construction projects that sit behind the current wave of AI infrastructure.
That focus matters because the constraint is real. AI is often discussed in terms of models and compute, but the foundation is physical. Data centers, semiconductor facilities, and energy infrastructure all need to be built before anything runs. Each of those projects requires insurance before work can begin. That step, historically, has been slow and manual.
The physical side of AI is where delays show up
Construction insurance underwriting was not designed for the pace at which these projects now move. Quotes can take weeks. Brokers spend time chasing updates across emails and calls. Information sits across disconnected systems. By the time a policy is priced, parts of the underlying risk may already be outdated.
This creates friction in a place that directly impacts timelines. If insurance lags, projects stall. That gap between speed of construction demand and speed of underwriting is where Shepherd positions itself.
From static paperwork to live project data
The shift Shepherd is making is relatively straightforward in concept. Instead of relying on static forms submitted at one point in time, they use live data pulled from construction platforms. That includes signals such as incident tracking, inspection activity, and on-site conditions.
This allows underwriting decisions to reflect what is actually happening on a project rather than what was reported weeks earlier. The immediate benefit is speed. Processes that previously stretched over weeks can be compressed significantly. More importantly, the data itself becomes more relevant.
Pricing risk based on how projects are run
Another important piece is how this affects pricing. Traditional models often group contractors into broad categories. Shepherd takes a more granular view by looking at how projects are executed in practice.
Contractors using better tools, maintaining stronger safety practices, and operating with more discipline can be priced differently. This introduces a feedback loop. Better operations can translate into better pricing, which creates an incentive to adopt stronger processes.
It also shifts underwriting from assumption-based to behavior-based. That is a meaningful change in how risk is evaluated.
Why this approach is gaining traction
The company’s growth reflects that this is not just a theoretical improvement. Strong revenue expansion and increasing coverage across large project portfolios suggest that the model resonates with both builders and insurance capacity providers.
The involvement of established insurers also signals something important. In a regulated space like insurance, distribution and capacity are not optional. New approaches still need to plug into existing structures. Shepherd appears to be doing that while changing how underwriting decisions are made.
The longer-term direction is clear. Moving more of the underwriting workflow toward automation, supported by continuous data rather than static submissions.
Key takeaways for fintech startups
There are a few practical observations worth calling out.
- Some of the most valuable opportunities sit in slow, operational layers that are easy to overlook
- Real-time data can materially change how risk is assessed when existing processes rely on outdated inputs
- Speed matters, but it becomes more powerful when paired with better decision quality
- Partnerships remain essential in regulated industries, especially where balance sheet capacity is involved
- Starting with a narrow, well-defined segment can help build depth before expanding into adjacent areas
If you are working on similar inefficiencies in fintech, there is often more room to build than it initially seems. If you want to explore how to turn that into a clear strategy, reach out.
