The commercial insurance sector has long promised that technology will streamline underwriting, yet the reality on the ground tells a different story. Ola Kolade, CEO of Underflow, argues that the industry has spent the last two decades polishing the exterior of processes while leaving the core manual labor untouched. By digitizing filing cabinets and creating prettier inboxes, vendors have merely given underwriters a more attractive workspace to perform the same tedious tasks. This fundamental misalignment explains why, despite billions poured into insurtech, the average underwriter still spends the majority of their day on data entry, attachment hunting, and follow‑up drafting rather than on risk assessment and pricing decisions. Kolade’s critique cuts to the heart of a systemic issue: automation efforts have targeted document management instead of the cognitive work that underwriters actually perform.
Supporting data from recent industry studies underscores the magnitude of the inefficiency. Capgemini’s 2024 World Property and Casualty Insurance Report revealed that 41 to 43 percent of a commercial underwriter’s schedule is consumed by administrative chores such as manual data entry and record keeping. Only about one‑third of their time remains for core underwriting functions like evaluating risk, calculating premiums, and exercising judgment. Accenture’s analysis paints an even starker picture, estimating that the typical underwriter devotes roughly 70 percent of their workday to non‑underwriting activities. When extrapolated across the sector, this waste translates into an estimated $85 billion to $160 billion of lost productivity over a five‑year horizon. These figures are not abstract; they are regularly cited by carriers, managing general agents, and wholesalers as a pressing bottleneck that inflates costs and slows response times.
Kolade’s own trajectory offers a unique lens through which to view the problem. A Canadian economics graduate from McGill, he began his career in investment banking at National Bank Financial, advising on technology‑focused transactions. He then moved to San Francisco to lead strategic finance at Faire, a leading wholesale marketplace, where he oversaw the upmarket business unit serving the platform’s largest retail clients. Prior to that, at Power, he structured $10 million of debt financing, built three‑statement financial models, and coordinated a strategic advisory board. His exposure to commercial insurance emerged indirectly through Faire, where he observed how thousands of independent businesses managed risk. The stark contrast between the sophisticated machine‑learning tools he used to match retailers with inventory in real time and the antiquated email‑thread, PDF‑attachment, manual re‑keying processes prevalent in insurance struck him as a fundamental mismatch between talent and tooling.
The crux of Kolade’s argument lies in distinguishing between automating the “container” and automating the “work.” He likens traditional insurtech solutions to upgrading a filing cabinet: the documents are stored more neatly, but the human still must open each file, read every page, transcribe data, and craft follow‑up messages by hand. In his view, true automation must remove those manual steps entirely, allowing the underwriter to receive a submission that is already interpreted, validated, and enriched. This shift from superficial digitization to deep cognitive assistance is what he believes will finally unlock the productivity gains the industry has chased for years.
Enter Aurora, Underflow’s autonomous underwriting assistant, which embodies this philosophy. Rather than building a faster version of the existing workflow, Aurora aims to eliminate the manual portions altogether. When a submission arrives via email, the system reads the message and every attached document, extracts all pertinent data, and assembles a unified, structured record. It then conducts a granular gap analysis that does more than flag missing information; it pinpoints exactly what is absent, why it matters, and how it relates to carrier requirements and the specific line of business under consideration.
Because Aurora operates directly within the existing email environment, brokers need not change their habits or learn a new portal. They continue to send submissions exactly as they always have; the difference lies in what happens after the email lands in the carrier’s inbox. The system automatically drafts and sends follow‑up requests to the broker, either under its own identity or through the carrier’s email address. If no response arrives within a preset window, Aurora issues a reminder, persisting until the file is complete. Once all required information is gathered, the package is routed to the assigned underwriter, ready for immediate review.
Beyond intake, Aurora adds layers of value that traditional tools lack. It writes preliminary risk assessments, highlights inconsistencies between documents (for example, a loss run that is two years short of carrier requirements or conflicting construction type data), and identifies potential coverage opportunities that might otherwise be overlooked. The platform also enriches submissions with external data pulled from public sources such as property records, business registries, and corporate profiles. This contextualization means the underwriter receives a file that is not merely complete but already analyzed, with relevant background information integrated, allowing them to focus on judgment calls rather than data gathering.
The urgency of Kolade’s vision is amplified by demographic pressures facing the industry. Roughly one quarter of the commercial insurance workforce is aged 55 or older, meaning a substantial reservoir of institutional knowledge—carrier appetite nuances, risk pattern recognition, and long‑standing relationship context—is poised to walk out the door with each retirement. Kolade notes that losing 20 to 30 years of experience per departure cannot be remedied by hiring alone; the mathematics simply do not support replacing seasoned judgment with novice labor. Instead, he advocates capturing the repeatable, rule‑based components of that expertise within a scalable system like Aurora, thereby freeing the remaining professionals to concentrate on the complex, human‑centric aspects of underwriting that truly require seasoned insight.
Insurance carriers have historically approached new technology with caution, and for good reason. Legacy systems are deeply entrenched, switching costs are high, and trust in unfamiliar platforms is earned slowly. Kolade anticipated these hurdles by designing Aurora to integrate seamlessly with the email workflow that already governs submission exchanges. By avoiding the need for brokers to adopt a new portal or for underwriters to learn a separate interface, the product reduces friction and respects the industry’s established communication channels. This approach lowers the barrier to entry and addresses a common critique of insurtech solutions: that they demand disruptive change rather than working within existing realities.
From a market perspective, Kolade’s framing resonates amid a broader shift toward AI‑augmented decision making across financial services. While many insurtech vendors have focused on improving document storage, OCR accuracy, or workflow routing, fewer have tackled the semantic understanding layer that Aurora targets. The success of large language models in extracting meaning from unstructured text suggests a fertile opportunity for tools that can interpret insurance documents in context. Competitors that continue to offer only incremental UI enhancements may find themselves outpaced as carriers seek solutions that demonstrably reduce underwriter workload and improve submission quality.
For insurance executives looking to evaluate or adopt technologies like Aurora, several practical steps can guide a successful implementation. First, conduct a baseline audit of how underwriters currently spend their time to quantify the manual burden and identify the specific tasks that consume the most effort. Second, involve frontline underwriters and brokers early in the design process to ensure the solution aligns with real‑world workflows and respects communication preferences. Third, prioritize vendors that can demonstrate deep contextual comprehension—such as gap analysis, risk‑assessment drafting, and external data enrichment—rather than those offering only superficial digitization. Fourth, plan a phased rollout that begins with a pilot group, measures key performance indicators like time‑saved per submission and error‑reduction rates, and scales based on verified outcomes. Finally, foster a culture of continuous learning where the AI system’s outputs are reviewed and used to refine both the technology and the underwriters’ judgment, creating a feedback loop that enhances overall performance over time.