Estimated reading time: 7 minutes
Cowbell’s new framework scores how much autonomy customers hand their AI systems. OMNI’s own settings are undescribed.
Cowbell’s underwriters accept their AI’s pricing nine times out of ten. The company launched that AI today. OMNI reads submissions, weighs risk signals, and returns priced non-admitted quotes in minutes. Cowbell reports 53% growth in new business since rollout.
The loss ratio on that business is another matter. Cyber Insurance News asked Cowbell to compare OMNI-underwritten results against the pre-OMNI book. “It’s too early to provide a meaningful loss ratio comparison,” a company spokesperson said. The portfolio has not matured enough to support the analysis.
Excess and surplus cyber business takes years to develop credible loss data. It is still worth noting which number was ready for launch day.
What Cowbell OMNI Actually Does
Cowbell announced OMNI on July 28, 2026. It targets a real problem. Submission volumes in small commercial specialty keep climbing. Per-account premiums do not support hand underwriting at that scale.
Cowbell’s answer is a layer of AI agents. They read submissions, pull risk signals, check appetite, and produce priced recommendations. Human underwriters approve or override. Eligible non-admitted quotes return in minutes rather than weeks.
The word “eligible” carries weight. Cowbell has not disclosed what share of submissions clear that gate.
Cowbell says OMNI “has been deployed incrementally, with underwriting serving as the first production use case.” Claims, cyber resilience, customer service, and product development follow.
Founder and CEO Jack Kudale framed the conditions driving the build. “Specialty lines are increasingly defined by bifurcation,” Kudale said. He cited pricing pressure, elevated loss costs, and litigation volatility.
Co-founder and Chief Product Officer Rajeev Gupta drew a line against the obvious comparison. “Insurance doesn’t need another chatbot,” Gupta said.
OMNI rests on three pillars. Intelligence covers the agents and the insurance-specific language models. Orchestration routes work and compress process time. Governance covers observability and performance measurement. Cowbell tracks that third pillar through Bellwether, an internal AI performance metric. No third party has audited it.
The Growth Number Arrives Early. The Performance Number Takes Years.
Cowbell answered follow-up questions on the 53% figure. It measures new business premium, not policy count. It dates from May 2026, when Cowbell put OMNI’s underwriting agents into production, measured against the period immediately preceding. The company has not defined that baseline period.
The figure covers every product OMNI’s underwriting agents support. That includes Prime One, which launched in the United States on April 21, weeks before the growth window opens. Cowbell says the number reflects the broader underwriting portfolio “rather than the impact of any single product.” The company has not published a breakout separating the two.
Both explanations remain live. Sharper underwriting grows new business premium. So does entering the middle market for the first time.
Market context sharpens the question. Cowbell’s own 2026 claims report showed US cyber premiums falling to $9.14 billion while claims rose 40%. Growth into a softening market with rising loss costs is precisely when loss ratio earns its keep.
The company has also promised the market discipline. Cowbell appointed John Botros as CFO in January with an explicit operating-profitability mandate.
Six Weeks, With An Asterisk
Cowbell says OMNI cut product deployment from roughly eight months to as little as six weeks. Asked whether that compared an admitted product against a non-admitted one, a spokesperson said no. “The eight-month and six-week figures refer to our product development lifecycle,” the spokesperson said.
Prime One is the stated proof point. It deserves a closer look.
Cowbell launched Prime One in Australia on February 6, 2026, on Zurich paper. The US version followed in April on Chaucer paper. At the US launch, Cowbell said the product adapted learnings from international deployments. Porting an established product across a border in six weeks is a genuine achievement. Building one from nothing in six weeks is a different claim.
A second gap sits inside the launch material. OMNI’s flagship application is high-volume SME excess and surplus business. Prime One serves companies with $250 million to $1 billion in revenue. Those are different books.
Cowbell has also not named the carrier writing the E&S paper behind OMNI-generated quotes.
Cowbell Now Scores AI Autonomy. OMNI’s Own Settings Are Undescribed.
Five days before the OMNI launch, Cowbell introduced three new AI Cowbell Factors. Gupta described the framework in a company blog post on July 23. The factors score observability, autonomy, and governance inside a policyholder’s AI environment. Cowbell now underwrites how much authority a customer hands its machines.
Gupta drew the distinction plainly. He separated an assistant that drafts language from an agent that “independently executes financial transactions, approves customer requests, or modifies production code.”
Cowbell applies that lens outward. It has not applied it publicly to OMNI.
The company says human underwriters hold final authority. It also says those underwriters accept roughly nine recommendations in ten. “Approximately 90% of OMNI’s pricing and coverage recommendations are accepted by our underwriters,” a spokesperson said. Cowbell has not defined acceptance. A quote bound at the recommended price and a quote bound after adjustment may both land in that number.
Gupta wrote about governance as an evidence-producing discipline. He argued that responsible AI “becomes measurable rather than merely aspirational.” Brokers placing OMNI-quoted business may reasonably ask to see that measurement.
Cowbell does not comment on how specific regulations apply to its business. That is stated company policy. The spokesperson pointed instead to the controls built into the system, citing “observability, human oversight, traceability, and continuous performance measurement.”
What Brokers And Risk Buyers Should Ask
Cowbell serves about 30,000 businesses worldwide. Those policyholders carry a combined $610 billion in insured revenue. The company has paid more than $275 million in claims since launch. Its distribution network passed 35,000 producers under Chief Commercial Officer Simon Hughes.
Cowbell also cites a global risk pool of more than 55 million entities. That figure counts entities Cowbell can model. It insures roughly 30,000 of them.
OMNI expands next across Europe and Asia-Pacific.
For brokers and risk buyers, five questions remain open. Which carrier writes the E&S paper. What share of submissions clear the eligibility gate. How Cowbell defines an accepted recommendation. What period the 53% is measured against. Whether any OMNI workflow executes without human sign-off.
Cowbell has built something substantial here. The numbers that will test it are still developing.
FAQ: Cowbell OMNI And AI Underwriting
OMNI is an AI-native decision intelligence system Cowbell announced on July 28, 2026. It runs underneath underwriting, claims, cybersecurity services, and product development. Its first major application handles high-volume, lower-premium excess and surplus business for small and mid-sized organizations.
Cowbell says it is too early to tell. A company spokesperson said the OMNI portfolio has not matured enough to support a meaningful comparison against the pre-OMNI book. Cowbell’s stated focus has been underwriting consistency, speed, and decision quality. Excess and surplus cyber business typically takes several years to develop credible loss data.
Cowbell says the figure measures new business premium following the May 2026 deployment of OMNI’s underwriting agents, compared against the period immediately preceding. The company has not defined the length of that baseline period. The figure covers all products OMNI’s underwriting agents support, including Prime One, which launched in the US in April 2026. Cowbell says the growth reflects the broader portfolio rather than any single product, and has not published a product-level breakout.
Cowbell says underwriters accept approximately 90% of OMNI’s pricing and coverage recommendations. The company has not published a definition of acceptance. It is unclear whether a quote bound after price adjustment counts as accepted or modified.
Cowbell says those figures describe its product development lifecycle. The company cites Prime One as evidence. Prime One launched in Australia in February 2026 on Zurich paper before reaching the US market in April on Chaucer paper. Cowbell said at the US launch that the product adapted learnings from international deployments.
Cowbell introduced three AI Cowbell Factors on July 23, 2026, expanding its risk-rating framework from eight factors to eleven. They measure observability, autonomy, and governance in a policyholder’s AI environment. The framework scores how much authority a customer gives its AI systems. Cowbell has not published equivalent detail about OMNI’s own autonomy settings.
Cowbell describes governance as one of OMNI’s three pillars, tracked through an internal performance measurement capability called Bellwether. A spokesperson cited observability, human oversight, traceability, and continuous performance measurement as built-in controls. Company policy bars comment on how specific regulations apply to the business. No third party has publicly audited Bellwether.
Related Cyber Insurance Posts
- Command Zero’s Throughline Turns Closed Cases Into Living Investigations
- Cyber Insurance Companies Face Loss Ratio Impact from Amazon Outage(Opens in a new browser tab)
- The Cyber Insurance Soft Market Is “Finding A Floor” – Will Every Carrier Survive?(Opens in a new browser tab)
- Do You Speak “Cyber?” For Brokers Who Don’t, Cowbell Wants to Help (Opens in a new browser tab)
- Underwriting Performance for US Cyber Insurance Improved Last Year: Reinsurance News(Opens in a new browser tab)