Sixfold

Our Testing Methodology

At Compare The AI, we take our responsibility to the insurance industry seriously. When evaluating an enterprise-grade underwriting automation platform like Sixfold AI, we cannot rely on marketing brochures or controlled demo environments. We need to see how the AI handles the messy, unstructured reality of actual insurance submissions. Our goal is to simulate the real-world chaos of an underwriter's inbox.

For our evaluation of Sixfold AI, we assembled a testing panel consisting of two former commercial Property & Casualty (P&C) underwriters with over 30 years of combined experience, one senior Life & Health (L&H) underwriter specializing in complex medical cases, and our core AI technical assessment team. We secured a full sandbox environment from Sixfold and, over a four-week period, fed it a diverse and challenging dataset of 150 anonymized, historical submissions. This dataset was intentionally complex, designed to push the platform to its limits, and included:

• P&C Submissions (100 cases): A broad mix of commercial applications ranging from straightforward Business Owner Policies (BOPs) and General Liability renewals to highly complex, multi-layered Cyber and specialty property risks. The data was intentionally messy, including a variety of formats: high-resolution ACORD 125 and 130 forms, poorly scanned ACORD forms with coffee stains, 500+ page loss runs in both Excel and locked PDF formats, contradictory supplementary documentation, and broker emails with critical information buried in the signature.

• L&H Submissions (50 cases): A curated set of individual life insurance applications designed to test the AI's clinical acumen. These were accompanied by dense Attending Physician Statements (APS) running over 200 pages, unstructured Electronic Health Records (EHRs) with physician-to-physician notes, and complex lab results showing fluctuating values over several years. We included cases with multiple comorbidities, such as a diabetic applicant with a history of cardiovascular issues and controlled hypertension.

Our testing protocol focused on three core pillars, with specific success metrics for each:

1. 1 Ingestion and Extraction Accuracy: How well does the AI parse unstructured data without human intervention? We measured the error rate in data extraction from low-resolution PDFs and handwritten notes. Our benchmark was to see if Sixfold could achieve over 98% accuracy on key data points (e.g., policy limits, applicant names, dates of diagnosis) without requiring a separate Optical Character Recognition (OCR) tool.

1. 1 Appetite Alignment and Risk Synthesis: Can the platform accurately learn and apply a specific, nuanced set of underwriting guidelines? We programmed Sixfold with three distinct, hypothetical risk appetites: an aggressive growth model focused on new business in emerging tech sectors, a highly conservative retention model for a legacy manufacturing book, and a balanced middle-market approach. We then measured how consistently the AI applied these rules to identical submissions, looking for any deviation in its risk scoring.

1. 1 Workflow Integration and Actionability: Does the tool actually save time and make underwriters more efficient? We established a human baseline by timing our expert underwriters as they processed a sample set of submissions manually. We then measured the time taken by the same underwriters to triage, assess, and generate referral narratives using Sixfold. The goal was to quantify the efficiency gains in a real-world scenario.

We spent four weeks actively using the platform, treating it as our primary underwriting workbench. We tweaked guidelines, challenged the AI agents with ambiguous edge cases, and manually verified every single citation the platform produced. The resulting review reflects our deep, hands-on experience with Sixfold as a daily underwriting companion.

What Is Sixfold AI?

Unlike generic Large Language Models (LLMs) or horizontal document extraction tools that simply perform OCR, Sixfold is not just reading text; it is interpreting risk. It acts as an "AI Brain" or a virtual underwriting analyst that sits between the submission inbox and the policy administration system. Its primary function is to ingest complex, multi-format submissions—loss runs, medical records, financial statements, broker emails—and synthesize that data into actionable insights that are directly aligned with a carrier's specific, and often unique, risk appetite.

The problem Sixfold solves is acute and universal. Across the industry, in both P&C and L&H, underwriters spend an estimated 40% to 60% of their time on low-value administrative tasks: re-keying data from PDFs into rating systems, hunting for specific clauses in 200-page documents, and manually summarizing findings for referrals. This "admin trap" is a massive drain on productivity. It reduces the time available for the high-value cognitive work that actually drives profitability: deep risk analysis, strategic broker relationship management, and proactive portfolio strategy. Sixfold aims to eliminate this manual heavy lifting, freeing underwriters to operate at the top of their license.

Key Features

Sixfold’s feature set is highly specialized, reflecting a deep understanding of the underwriter's daily workflow. It is not a collection of disparate tools, but a cohesive platform designed to manage the submission lifecycle from intake to decision. We broke down its capabilities into the following core areas:

1. Appetite-Aligned Submission Triage

The front door of the underwriting process is often a chaotic, first-in-first-out inbox. Sixfold brings intelligent order to this chaos. It automatically ingests and scores every incoming submission on a scale of 1 to 5 based on how well it fits the carrier's pre-defined underwriting guidelines. In our testing, this feature was transformative. Instead of working through an inbox chronologically, our underwriters could instantly see which submissions were high-priority "in-appetite" risks (scoring a 4 or 5) and which were immediate declines or required minimal touch (scoring a 1). The AI doesn't just provide a score; it provides the why, highlighting the specific data points that drove the classification, such as a high-hazard class code or a history of severe losses.

2. Deep Document Ingestion and Risk Assessment

This is Sixfold's core strength. It eliminates the need for third-party data extraction tools by natively handling messy, unstructured documents in any format. For P&C, this means parsing multi-year loss runs with inconsistent formatting and reading scanned ACORD forms with handwritten notes. For L&H, it means synthesizing hundreds of pages of Attending Physician Statements (APS) and lab results into a coherent clinical narrative. The platform generates tailored "Risk Signals" that are specific to the line of business. For example, if a commercial property submission includes a building with an outdated roof and no fire suppression system, Sixfold flags these as distinct negative signals, citing the exact page and paragraph in the inspection report for each. In L&H, it connects medications to diagnoses, building a complete clinical picture and flagging critical discrepancies between what the applicant disclosed and what the medical records show.

3. Transparent Citations and Traceability

One of the biggest hurdles to AI adoption in the highly regulated insurance industry is the "black box" problem—underwriters and regulators cannot trust a decision they cannot verify. Sixfold solves this brilliantly with its best-in-class in-line citations. Every insight, summary, and risk signal generated by the AI includes a direct, clickable link back to the source document. In our testing, clicking a citation instantly opened the original PDF, highlighting the exact sentence or data point the AI used to make its assessment. This complete traceability is crucial for building user trust, ensuring regulatory compliance, and facilitating internal audits. It turns the AI from a mysterious oracle into a verifiable assistant.

4. Autonomous Action Agents

Moving beyond simple analysis, Sixfold deploys "Action Agents" to handle downstream tasks and clear bottlenecks. These are pre-trained AI modules that execute specific underwriting workflows. For example, if a submission falls outside an underwriter’s authority limit, the Referral Agent automatically drafts a detailed referral email to a senior underwriter, summarizing the risk, explaining the specific referral triggers, and attaching all the relevant documentation. The Research Agent can conduct automated web research to enrich the submission data, checking for adverse media, verifying business classifications (SIC, NAICS), or looking for recent safety violations.

5. Flexible Integration and Workflow

Sixfold is designed to work where the underwriter works, not force them into a new, alien environment. It offers a clean, intuitive standalone web interface but also provides a robust and well-documented set of APIs to integrate directly into existing underwriting workbenches, CRM systems (like Salesforce), or major policy administration platforms (like Guidewire, Duck Creek, or Majesco). This flexibility allows carriers to embed Sixfold's intelligence into their established workflows, minimizing disruption and accelerating adoption.

Performance in Testing

Our four-week testing period revealed a platform that is remarkably adept at its core mission, though it demands a thoughtful implementation to achieve its full potential.

What Worked Exceptionally Well

The L&H Medical Synthesis: We were genuinely astounded by Sixfold's ability to handle complex L&H submissions. We created a test case for a fictional 58-year-old male applicant with a history of well-managed Type 2 Diabetes, controlled hypertension, and a family history of heart disease. We fed the system a 300-page, highly disorganized medical file containing a mix of typed and handwritten doctor's notes, repetitive lab results, and a specialist's report from five years prior. Within minutes, Sixfold produced a coherent, chronological health narrative. It correctly identified the applicant's HbA1c levels over time, noted the consistency of his medication regimen, and flagged a previously undisclosed minor cardiac event from the specialist's report that was only mentioned in passing. The time savings here are not incremental; they are exponential. What would take a human underwriter 2-3 hours of painstaking review took the AI less than five minutes, with every finding cited.

P&C Triage and Risk Signal Accuracy: We tested the P&C capabilities with a submission for a fictional construction company, "Mirage West Construction," specializing in roofing. We programmed our hypothetical "conservative" risk appetite to be wary of contractors with more than 50% of their operations in roofing. The submission documents buried this detail on page 47 of a supplementary questionnaire. Sixfold not only flagged it instantly but also created a high-priority Risk Signal: "Exceeds authority threshold of 50% roofing operations." It ruthlessly and accurately filtered out submissions that violated our knockout rules, catching subtle exclusions in loss runs that our human testers initially missed during a rapid manual review. The 1-5 scoring system proved highly reliable; we rarely disagreed with a submission that Sixfold scored as a 1 (clear decline) or a 5 (clear quote).

The Citation Engine: The transparent citations are the killer feature that underpins the entire platform's value. In our testing, the AI occasionally misinterpreted a highly ambiguous, jargon-filled clause in a commercial lease agreement related to liability. However, because the citation linked directly to the source text, our human underwriter instantly spotted the AI's error, understood the context, and made the correct judgment call. The AI is a powerful assistant, but the human remains the final arbiter, and Sixfold's UI perfectly supports this essential human-in-the-loop dynamic.

Where We Found Friction

Initial Guideline Ingestion: While Sixfold claims to easily ingest underwriting guidelines, we found that the quality of the AI's output is heavily dependent on the clarity and structure of the carrier's internal documentation. If your underwriting manuals are a collection of vague, contradictory, or outdated Word documents, Sixfold will struggle to build a coherent model. We had to spend significant time refining our hypothetical guidelines into a more structured format to ensure the AI interpreted them correctly. It requires a carrier to have their own house in order before the AI can be truly effective. This is not a magic wand for a messy process.

Complex Multi-Layered P&C Risks: While excellent at standard and mid-market commercial lines, Sixfold occasionally struggled to synthesize a holistic narrative for highly complex, multi-layered specialty risks, such as a global energy portfolio with dozens of distinct legal entities, varying international coverage limits, and complex reinsurance structures. It could extract all the individual data points correctly, but the synthesized narrative sometimes lacked the nuanced, big-picture perspective that a 30-year veteran senior underwriter would provide.

Who Should Use Sixfold AI?

Sixfold is not a tool for small retail brokerages or individual independent agents. It is a powerful, enterprise-grade solution built for carriers and large MGAs that deal with significant submission volume and document complexity.

Ideal Users:

• Commercial P&C Carriers: Insurers handling high volumes of mid-market to large commercial submissions (especially in lines like General Liability, Commercial Property, Workers' Comp, and Cyber) where manual document review is a major bottleneck.
• Life & Health Insurers: Carriers processing complex individual life, disability, or long-term care applications that require extensive and time-consuming medical record review. This is a key differentiator for Sixfold.
• Managing General Agents (MGAs): Large MGAs with delegated underwriting authority who need to process submissions rapidly and demonstrate consistent guideline adherence to their carrier partners.
• Reinsurers: Companies needing to audit large portfolios of bound policies from their cedents to ensure compliance with treaty terms and identify risk accumulation.

Specific Roles That Benefit Most:

• Chief Underwriting Officer (CUO): To gain a real-time, data-driven view of portfolio health, ensure consistent guideline application across the organization, and strategically allocate underwriting resources.
• Underwriting Managers/Directors: To monitor team capacity, reduce quote turnaround times, and use analytics to identify profitable niches and problematic broker submissions.
• Senior Underwriters: To delegate the administrative burden of data extraction and focus their expertise on complex referrals, broker negotiations, and strategic risk selection.
• Underwriting Assistants/Analysts: To evolve their role from manual data entry to validating AI-generated insights, becoming a crucial part of the human-in-the-loop quality control process.

Sixfold AI vs The Competition

The AI underwriting space is heating up rapidly, but different platforms are solving different parts of the problem. Here is how Sixfold compares to its primary vertical competitors:

Pros & Cons

Pros:

• Exceptional Unstructured Document Parsing: Handles messy, multi-format P&C and L&H documents with remarkable accuracy, effectively eliminating the need for separate OCR or data extraction tools.
• Game-Changing L&H Medical Synthesis: Transforms hundreds of pages of chaotic medical records into a clear, actionable clinical summary in minutes, creating a massive efficiency gain.
• Trust-Building Transparent Citations: The in-line citation feature is best-in-class and builds immediate trust by linking every single AI insight directly back to the source document.
• Truly Customizable Risk Appetite: The platform adapts to your specific, nuanced underwriting guidelines rather than forcing you into a generic, one-size-fits-all risk model.
• Effective Action Agents: The pre-trained AI agents are highly effective at automating tedious downstream tasks like drafting referral emails and conducting basic web research.

Cons:

• Requires Clean, Structured Guidelines: The AI is only as good as the underwriting manuals you feed it; vague, outdated, or contradictory guidelines will lead to poor AI performance and a frustrating implementation.
• Complex and Resource-Intensive Implementation: This is not a plug-and-play solution. It requires a significant upfront investment in time and resources from your best senior underwriters and IT staff to properly train and calibrate the AI.
• Struggles with Extreme Complexity: While improving, it can still oversimplify the narrative on highly complex, multi-layered global specialty risks that require a deep, holistic understanding of geopolitical or systemic risks.

Compare The AI Verdict