Nettle, a London company that automates commercial insurance risk assessments, has raised $4.8M in an oversubscribed seed round led by MTech, taking the total it has raised to $6.8M. Project A Ventures, Sure Valley Ventures, Portfolio Ventures and Ventures Together joined the round alongside several angel investors, according to the company's announcement, with Project A having also led the £1.45M pre-seed in April 2025.
The work the company automates sits at a point in insurance that policyholders never see. Before an insurer writes a large commercial policy on a factory, a warehouse or a construction site, a risk engineer visits the premises, inspects the structure, the machinery and the fire protection, and writes a report that underwriters price from.
That report has become the constraint. Nettle puts inspection backlogs across the industry at up to six months, and says risk engineers spend roughly 90% of their time on analysis and writing rather than on the site visits themselves.
Its platform works at both ends of that job. Before anyone travels, it reads external data to flag likely hazards and produces an inspection guide for the specific site, and during the visit it takes in images, video, audio notes and documents from a phone, then assembles the risk report afterwards.
The numbers the company publishes describe a large change to a slow process. Nettle says time per project falls about 80%, which it presents as four times the output per engineer, and that engineers using the system complete five times more inspections. Allianz is the named customer, including a pilot at Allianz Türkiye announced in April, where the company reported inspections running up to three times faster. The American insurer Brotherhood Mutual also uses the platform, which covers property, liability, construction and workers' compensation lines.
Jack Miller and Katya Lait founded the company, both out of QuantumBlack, McKinsey's AI arm. Miller, the chief executive, was a product manager there across insurance deployments, while Lait, the chief technology officer, built the firm's first multi-agent generative AI product for financial services.
That architecture carried over. Nettle runs a multi-agent system that distributes work across specialised agents, caches results, and can be deployed on a customer's own infrastructure, which is how it answers insurers whose underwriting data cannot leave their premises. "Underwriters and brokers have kind of missed the SaaS wave and have jumped straight to AI," Lait has said of the market the company is selling into.
The Backlog Is A Staffing Problem
The queue Nettle describes exists because the work is manual and the people who do it are scarce. Commercial loss control surveys are conducted on foot by consultants who document operations, premises, protection and special hazards for property, liability and workers' compensation policies. The loss control team at the claims firm Davies alone carried out 119,976 on-site surveys across all 50 American states in 2025.
Each of those requires an experienced engineer, and the supply is tightening. Nettle estimates that 40% of risk engineers could retire by 2030, a figure it has used since its pre-seed round without publishing the source behind it.
Project A, its earliest backer, framed the same problem as a capacity ceiling rather than a cost one. An insurer that cannot get sites surveyed cannot write the policies, which caps how fast its commercial book can grow regardless of how much appetite the underwriters have. The company says its proof-of-concept partner used the platform to support twice the underwriting growth, though that claim, like the speed figures, comes from Nettle rather than from the insurer.
Insurtech Money Now Goes Only To AI
The round lands in a funding market that has narrowed to a single theme. Gallagher Re counted $2.44B of insurtech funding in the second quarter of 2026, of which $2.42B, or 99.1%, went to companies built around AI, while reinsurers and insurers themselves backed 27 technology investments in the period.
The early stage was the weaker part of that picture. Seed and pre-seed deals raised $264.19M across 54 transactions, down 51.8% from $548M the previous quarter, which makes an oversubscribed seed round a less common outcome this year than the headline total suggests.
What the capital is chasing is visible in where it has already gone. Insurance has absorbed AI fastest in the parts of the business that generate documents, and risk reports are documents produced at considerable expense by expensive people.
The harder half of the market is the part where the output drives a decision. A model that drafts a report leaves the pricing judgement with an underwriter, and the step from reading conditions to acting on what the reading shows is where most industrial AI deployments have stopped short.
Regulators Are Watching AI Inspection
Supervisors in the United States are already looking at automated property assessment, though their attention has so far fallen on the personal lines rather than the commercial ones. Research by the National Association of Insurance Commissioners found that more than 70% of homeowners insurers were using, developing or exploring AI, with nearly half applying it in underwriting decisions, and its model bulletin requires insurers to run governance programmes, monitor outcomes, test for bias, validate data quality and oversee third-party vendors.
California has moved furthest. Assembly Bill 1559 would require insurers to tell homeowners when aerial imagery has been collected and to hand over the images when they are used to cancel or decline cover, so the accuracy can be disputed. The complaints behind that bill point at a specific failure: policyholders have reported adverse decisions based on outdated imagery, including roofs flagged as damaged after repairs were finished, shadows read as defects and trees marked long after removal.
Nettle sits on different ground. Its evidence comes from an engineer standing on the site rather than from a satellite pass, and the report it produces goes to an underwriter rather than straight to a renewal decision.
The governance questions follow the technology regardless. An insurer running a multi-agent system over its own inspection archive still has to document what the system touched, how it reached a conclusion and where the agent's authority ends, which is the same paperwork the NAIC bulletin asks for. The competitive field reflects that split, with Cape Analytics, Zesty.ai, Verisk, Nearmap and Cotality supplying imagery-based property assessment largely to personal lines, while the commercial survey remains a visit by a person with a clipboard and, increasingly, a phone.
A Small Company Holding Large Accounts
What Nettle has assembled is unusual for its stage. Allianz and Brotherhood Mutual are substantial institutions, and the company is working towards 12 full-time employees across London and New York. That ratio cuts both ways: named insurers give a seed-stage company credibility that marketing cannot buy, and they also arrive with procurement, security review and audit obligations that consume a small engineering team.
The figures available are the company's own. A platform that claims five times the inspection throughput and 80% less time per project is describing an operational change an insurer can measure precisely, and none of the insurers involved has published those measurements.
What the round establishes is narrower. An AI company selling into commercial underwriting has raised money in a quarter when early-stage insurtech funding halved, on the strength of a bottleneck the industry agrees exists and a set of results it has not yet verified in public.
