To grow, insurers need to increase policy count while simultaneously creating operating leverage, decoupling growth from operating expenses.
I believe there are four things insurers should keep in mind to maximize the opportunity presented by AI. We often describe these as phases in a maturation process.
Phase 1: Digitizing single communications to your view of risk
A root cause of much of the friction and inefficiency that exists in commercial insurance workflows is the vast amount of unstructured data that passes between insured, broker, and insurer. Typically emailed, often in the form of old policy documents, partially completed forms, and other supplemental information, this data must be re-entered by underwriting organizations before underwriters can act.
Digitization solves this by turning unstructured information into actionable data. The most effective solutions not only turn unstructured information into data but also use LLMs to specifically digitize this information into the insurer's view of risk. The most effective solutions are 'schema-driven', rather than digitizing to a standardized list of data fields, they digitize information into the specific data that each insurer needs to evaluate risk.
In short, effective digitization uses LLMs to eliminate manual data transcription while protecting the
Phase 2: Automating full risk capture
Digitizing a single email or form can create efficiency, but the underwriting process is rarely linear. The submission-creation process involves multiple communications, occurs across multiple channels, and happens over multiple points in time.
The best digitization solutions, therefore, not only digitize a single communication, but digitize a complete view of risk. Leading solutions accomplish this by linking multiple communications across multiple sources and multiple points in time, more accurately reflecting how underwriters actually prepare their risk for evaluation.
Phase 3: Automating end-to-end workflows
One and two address the data problem — turning unstructured submissions into a complete, insurer-specific view of risk. But even with perfect data, significant non-value-additive work remains: routing submissions, triaging follow-up, assigning underwriters, and coordinating across systems. Three is where AI moves beyond data capture and begins executing these workflow steps end-to-end, keeping the underwriter in the loop for judgment calls while eliminating the coordination work that prevents them from evaluating more risk.
Effective end-to-end orchestration capabilities should follow modern design principles, facilitating relatively modular technology stacks. Leading agentic platforms accomplish this through the concepts of skills and tools. For example, imagine a markdown file that describes how an underwriter should be assigned and can be updated when something changes. As a submission is received, the system can reference this markdown and assign an underwriter. The same concept could be applied to an appetite guide or an underwriting manual. Tools are more powerful extensions of this capability, in that they can interact with other agents or technology solutions.
One economically valuable outcome of this capability is an increase in straight-through-processing (STP). By combining AI-powered digitization and end-to-end workflow orchestration with distribution platforms that connect to insurers' commercial rating APIs, insurers can accept an email from a broker, digitize the communication, automate follow-up until the submission is complete, and then route the submission to their rating APIs to return a quote.
Phase 4: Embedding in broker workflows
The first three phases of this maturation cycle are fundamentally about what happens inside the insurer: building a data foundation, completing the picture of risk across every communication, and executing the workflow steps that surround underwriting. Each phase compounds the value of the one before it, and together they represent a meaningful step-change in how efficiently an insurer can operate. The reason a fourth phase exists at all is that commercial insurance is not a single-party process. It is built on a distribution relationship — and no amount of internal automation removes the friction that originates on the broker's side of that relationship.
The friction is highest at the handoff: incomplete submissions, status uncertainty, and misaligned appetite signals. By designing AI-powered workflows with the broker in mind, insurers can give brokers real-time visibility into submission status, surface the exact data requirements the insurer needs, and proactively signal their appetite. This dramatically compounds improvements to operating efficiency, growth, and portfolio construction that AI provides inside an insurer's own workflows. Taken to its logical end, insurers that can partner with brokers to access policy data already held in broker management systems open the door to an experience, where the broker initiates the conversation rather than waiting for the broker to package and send it.
Brokers place risk with the insurers who make it easiest. Insurers that can operationalize all four phases would not just be more efficient — they are competitively differentiated, attracting better submissions and building broker relationships that are structurally harder to displace.
The reality is leading insurers are recognizing opportunity. A market characterized by deliberation and risk aversion is moving quickly to put this technology to use. Previous excuses to digital transformation, like siloed data and change management, are less meaningful in a world where solutions can integrate with existing downstream systems.
Those that can successfully execute this transformation and seize this moment will unlock meaningful growth without needing to scale their operating expenses or sacrifice their unique view of risk.
