Last week, I sat across from the CIO of a small regional carrier. They have an enterprise LLM license. Employees use it daily for drafting emails, summarizing documents, and answering questions they used to Google. No AI is in production. No workflows have changed. No broker has noticed a difference. But the CEO has issued the mandate: be known for AI in three years.
They are not an outlier. They are the median.
The gap between an LLM license and an organization genuinely known for AI is not primarily a technology gap. It is a sequencing gap, a culture gap, and a partnership gap. If any of that sounds like your organization, this is the playbook for you.
Start by clearing the AI hangover. Most organizations running an enterprise LLM for 12 to 18 months have more ungoverned AI activity than the CIO realizes. Well-intentioned teams have built prompt chains, automated exports, and decision shortcuts that nobody formally sanctioned. Frame the discovery as amnesty, not an audit with the goal to surface what exists, formalize what works, and retire what creates regulatory or operational exposure before it becomes a problem.
Then do the data work. The LLM is sitting on top of an unresolved data problem. Before any agentic workflow can function reliably, before any underwriting model can be trusted, the carrier needs an honest audit of what its data environment contains including what is clean, what is accessible, and what would silently undermine any model trained on it. This is not exciting work. It determines whether everything that follows is built on rock or sand.
The third investment is an advisory partner with genuine insurance domain expertise, not a technology vendor and not a generalist integrator. This is someone who has worked inside carriers and understands how underwriting decisions actually get made, and knows what AI has worked in comparable organizations versus what looked good in a demo. Their role in 2027 is not to build anything. It is to help the carrier understand its starting point, identify the use cases worth pursuing, and prepare the organization for the platform and deployment decisions that come in 2028.
Someone in the C-suite needs to own AI strategy with genuine authority across underwriting, claims, technology, and compliance. Not a task force. Not a steering committee. A person with a mandate.
The next step is to learn how to experiment well because there are two kinds of AI experimentation in the market right now. The bad kind is the seven-figure or more partnership with a brand-name vendor announced at a conference, designed as much for the press release as the result. It produces a nine-month pilot, a slide deck, and nothing in production. The good kind starts by asking the people closest to the work where AI could make their processes better. This looks like the underwriter re-keying submission data every morning, the claims handler spending an hour finding coverage precedent, or the analyst producing the same report seventeen times a year. These people know exactly where the friction is. They have simply never been asked.
Smart experimentation looks like AI build days leveraging outside practitioners who show rather than tell, and celebrating the experiment that produced a useful insight even when it didn’t reach production. Every experiment has a defined question, a timeframe, and a decision at the end to build, adapt, or stop. The right metric is workflows changed, not seats purchased.
The carrier enters 2028 with a clear picture of its data, a year of honest experimentation, validated use cases, and an advisory relationship mature enough to accelerate what comes next. That foundation is what makes the platform decision a strategic choice rather than a leap of faith.
In 2028, select and implement an enterprise workflow and orchestration platform, one with pre-built insurance process frameworks, integrated governance tooling, and the ability to connect data, automation, and oversight within a single architecture. A small carrier cannot afford to assemble this from point solutions. The operational and integration cost of a fragmented approach will consume the budget before anything reaches production. Because the carrier now knows what it is building, informed by 2027’s data work and experimentation, the selection is grounded rather than speculative.
The first broker-visible outcome is submission turnaround. An AI-assisted underwriting workflow built on clean data can reduce time-to-quote from days to hours. The second is appetite clarity: AI-assisted triage that tells a broker earlier and more precisely whether a risk fits, what additional information helps, and what the carrier’s likely response will be. Brokers experience transparency as respect for their time.
Governance becomes an enabler rather than an obstacle for carriers that laid the groundwork in 2027. AI orchestration, managing multiple AI systems, data flows, and automated decisions within a coherent architecture, is what separates carriers running a collection of use cases from carriers running an AI-enabled operation.
Organizations should focus on a shift from making experimentation safe to making AI capability a source of professional pride. The underwriters who work best with AI tools need to be visibly rewarded not as a special category, but as the new definition of excellent underwriting. Bring the broker-facing team into the AI roadmap explicitly; they know exactly where the remaining friction is. Hire and train going forward for AI fluency presenting as people comfortable working with AI outputs, questioning them when something looks wrong, and improving the workflow based on what they observe.
By 2029, the CEO’s ask is answerable with evidence from the broker relationship. This looks like risks written with confidence that competitors decline and talent arriving because the reputation precedes the organization.The culture built in 2027 and 2028 has become the moat that technology alone cannot replicate.
The market is reorganizing around carriers with mature AI infrastructure. This looks like AI-native MGAs, operating on delegated authority without legacy system constraints, with commoditized underwriting for standard risks. Carriers that built AI capability will find MGA relationships repricing in their favor to include better delegation terms and partnership structures unavailable to carriers whose data and workflow maturity cannot support them. Those without AI capability face a binary choice: invest under pressure or accept a capacity provider role behind MGAs that own the customer relationship and the model.
By 2029, reinsurance treaty conversations will incorporate carrier AI capability as an underwriting variable. Behavioral pricing models, agentic claims infrastructure, and governance documentation will be rewarded in cost of reinsurance. That advantage compounds over multiple treaty cycles.
The culture achievement of 2029 is that AI is no longer a change management initiative. It is how the organization works. New hires assume it. Brokers expect it. The board measures it. The CIO’s job has shifted from evangelist to architect, designing the systems that ensure the carrier learns and adapts faster than the market. Learning velocity is the sustainable competitive advantage.
The three-year timeline is achievable but not forgiving of wrong turns. Eighteen months lost to a failed point solution, a platform selected before the use cases were understood, or a governance framework built under regulatory pressure cannot be recovered within the window.
Clear the AI hangover before building on top of it. Get the data right before selecting the platform. Establish the advisory relationship before committing to an architecture. Build the culture alongside everything else, not after it.
The bottom line is that the broker who notices the difference in 2029 doesn’t know what platform is running underneath. They know you quote faster, write risks with more confidence, and feel like a different kind of organization to work with.
The carriers that close the gap fastest find partners capable of working across the full stack, data, workflow deployment, and governance, rather than assembling point solutions from separate vendors. Carriers who choose a platform with the depth and configurability will grow rather than constrain their ambition. And while I’ve set all of this up as a roadmap for mid and small carriers, large carriers who want to be successful will be following this same playbook.
Bill Devine is Co-Founder and Managing Partner at Naitiv Partners, an AI-native consultancy focused on insurance and financial services transformation. He previously served as a senior executive at Travelers Insurance and as a Director at ACORD Standards.
.svg){kind=icon}
