The specialty insurance market has never been more dynamic, or more demanding. U.S. domestic excess and surplus lines premiums grew 13% year-over-year to $98.2 billion in 2024, representing the sixth consecutive year of double-digit growth. For a firm like Ryan Specialty, which operates across wholesale brokerage, managing general underwriters, and specialty programs spanning energy, healthcare, construction, marine, transportation, and dozens of other complex risk niches, that growth creates as much operational pressure as it does opportunity.
The problem is structural. Submission volumes in mid-market and E&S lines are up 35% year-over-year while underwriter capacity is flat or declining due to retirements and burnout, with over 50% of underwriter time still consumed by low-value tasks like data extraction and email triage. Supply and demand are misaligned, and the gap is widening. The question facing Ryan Specialty and the broader specialty market is not whether AI can help, but whether the industry can deploy AI at the depth and scale that the problem actually requires.
Most early AI deployments in insurance have been answering questions. They extract data, classify submissions, and surface summaries. That is useful, but it is not transformative. The real opportunity sits one level higher: AI that can reason about the full context of a risk, act autonomously across complex multi-step workflows, and drive decisions that close the gap between submission volume and underwriting capacity. That is the architecture Datafi was built to deliver.
The specialty insurance capacity crisis cannot be solved by AI that merely answers questions. Closing the gap between surging submission volumes and flat underwriter headcount requires autonomous AI agents that reason over full business context and act across end-to-end workflows.
From Marketplace Friction to Marketplace Intelligence
Ryan Specialty’s marketplace model, connecting retail brokers to specialized MGUs across dozens of risk niches, is fundamentally a supply and demand matching problem operating at enormous scale. Brokers submit risks. Underwriters assess appetite, price, and bind. Every step in that chain involves data retrieval, risk scoring, relationship context, portfolio positioning, and regulatory compliance, most of it handled manually today.
An AI operating system designed for this environment does not replace that chain. It makes the chain intelligent. With Datafi, the full data ecosystem of a specialty insurance operation becomes accessible to AI agents and workflows in real time. Submission data, loss history, third-party risk intelligence, carrier appetite rules, broker relationship context, treaty terms, and portfolio exposure are no longer siloed assets that underwriters must manually assemble. They become a unified context layer that every AI agent and workflow can operate against.
In the near future, nearly all customer onboarding functions in insurance could be delivered through AI multi-agent systems, where an intake agent ingests information, a risk profiling agent builds a comprehensive risk profile using existing underwriting guidelines, a pricing and product agent automatically prices the case, a compliance agent reviews for regulatory compliance, and a decision orchestrator agent aggregates input to determine if a policy can be automatically approved or escalated to a human underwriter. That is not a vision. That is the architecture Datafi enables today, applied to the specialty market.
What a Vertically Integrated AI Stack Changes
The challenge with point solutions in specialty insurance is the same challenge that afflicts enterprise AI broadly: fragmentation. An AI tool that scores submissions cannot see the carrier’s live portfolio exposure. A chatbot that answers broker questions cannot access the treaty structure that determines what the underwriter can actually bind. A workflow automation tool that routes emails cannot reason about whether a risk fits the MGU’s current appetite given recent loss development.
At Datafi, our view is grounded in what we have seen work and what has failed across complex data environments. LLMs need the full context of the business, access to the complete data ecosystem, and the ability to function in fully autonomous roles to learn and solve hard business problems. That is not a positioning statement. It is an architectural requirement. Without it, AI answers questions. With it, AI solves problems.
For Ryan Specialty, that distinction matters enormously. Specialty risk engineering tools that leverage AI are already cutting quoting times from more than one month to just days, while commercial and specialty P&C models incorporating predictive win rates now deliver quotes in one to two hours instead of two to three days. But those results depend on the AI having access to the right data, in the right form, with the right governance controls applied. A vertically integrated stack makes that possible. A collection of point solutions does not.
The Datafi Operating System for AI delivers unified data access across structured and unstructured sources, embedded governance and policy controls, autonomous agent and workflow capabilities, and a Chat UI designed for non-technical users. That last element is not cosmetic. Specialty insurance employs a broad range of professionals: underwriters, actuaries, claims handlers, compliance officers, relationship managers, and operations staff. If the AI layer is only accessible to data teams, the majority of value remains trapped. Datafi’s Chat UI was built explicitly to make AI capacity available to every employee, regardless of technical background, across every workflow that matters.
Operational Impact Across the Specialty Value Chain
Consider what this architecture enables across Ryan Specialty’s core operating areas.
In submission triage and underwriting, AI agents continuously ingest broker submissions, cross-reference against live portfolio exposure and carrier appetite, score each risk for fit and win probability, and route prioritized submissions to the right underwriter with the supporting context already assembled. Underwriters spend their time on judgment, not data assembly.
In risk engineering and pricing, predictive maintenance and asset management models built on historical loss data, third-party inspection records, and real-time IoT signals allow specialty lines underwriters in energy, construction, transportation, and manufacturing to price physical asset risk with a precision that manual processes cannot approach. The AI does not replace the underwriter’s expertise; it gives that expertise far better information to work with.
In broker relationship management, AI agents monitor submission flow, quote-to-bind ratios, declination patterns, and communication history to surface relationship insights and intervention opportunities that would otherwise require a dedicated analyst to track. Experienced producers become significantly more effective without adding headcount.
In operations and compliance, workflow agents handle document processing, regulatory filing, audit trail creation, and coverage verification, freeing operations staff for exceptions and complex cases rather than routine administration.
The Contextual Layer Is the Competitive Moat
MGUs that can consolidate, enrich, and protect proprietary data will become indispensable partners to both brokers and carriers, feeding better risk insights back into the ecosystem, and those that combine strong relationships with data ownership and advanced AI use will differentiate themselves most sharply.
That proprietary data layer, continuously enriched by every submission, bind decision, loss outcome, and broker interaction, is what Datafi calls the contextual layer. It is the accumulated intelligence of the business made available to every AI agent running on the platform. It is how an AI operating system learns the nuances of a specific risk niche, a specific book of business, a specific carrier appetite, rather than operating from generic training alone.
The global AI-powered insurance underwriting market is projected to grow from $2.85 billion in 2024 to $674 billion by 2034, expanding at a CAGR of 44.7%. That growth will concentrate in organizations that deploy AI at the architectural level, not the feature level.
That growth is not going to accrue uniformly. It will concentrate in organizations that deploy AI at the architectural level, not the feature level. For Ryan Specialty, the opportunity is to make its platform, its data, its relationships, and its underwriting expertise the foundation of an AI operating system that is genuinely impossible to replicate. The contextual layer that Datafi builds for each customer becomes a durable competitive moat, not a commodity capability that any competitor can license from the same vendor.
The specialty insurance market is complex by design. It exists because standard markets cannot price or structure certain risks with adequate precision. AI that truly understands the full context of that complexity, and can act autonomously within it, does not threaten that value proposition. It amplifies it.
Datafi (datafi.co) is the Business AI Operating System for the enterprise, delivering vertically integrated data access, autonomous agent and workflow capabilities, embedded governance, and a Chat UI designed for every employee.
