Proprietary Knowledge Is the New Moat
Most conversations about enterprise AI begin with the choice of model, yet that choice matters less with each release cycle. Zywave CTO Doug Marquis sees frontier and open-source options performing within a narrow range of each other on most tasks, and our own project work points the same way. The contest shifts to whoever holds the richest domain knowledge.
Zywave has built software for insurance brokers and carriers since 1995 and serves more than 15,000 customers, including all of the top 100 US firms. Its tools cover lead generation, marketing, and quoting, which gives it a close view of how business is won. According to Marquis, that vantage point has produced the industry’s largest archive of insurance losses, along with years of records on what buyers choose and what they pay. A general-purpose AI vendor has no comparable insurance data to draw on, so tools built for the industry tend to deliver sharper results in the sales and quoting work that brings in new business.
The Super Broker Takes Shape
Insurance agencies are being asked to grow faster with fewer people. Private equity firms have invested heavily in the sector on the expectation that agencies will expand their books by 10% or more per year, yet experienced brokers are retiring faster than new ones enter the profession. Hiring alone cannot close that gap, so Zywave built its Apex platform, whose Producer and Advisor agents take over the front half of the sales cycle: defining an ideal customer profile, finding the right buyer within a target company, drafting outreach tailored to that buyer’s circumstances, and laying the groundwork for a proposal.
The outcome is the AI super broker – a professional freed from administrative work to lead the consultative conversation in which coverage, risk, and cost are weighed. Early results show open rates and meetings booked running two to three times above conventional outreach baselines, four to five in some niches.
Intelligence Leaves the Screen
Insight is also traveling beyond the software that produces it, thanks to MCP, short for Model Context Protocol. It is a standard that allows an AI assistant to access external systems, such as an insurer’s loss history, and use that information to answer a question.
With MCP servers in place, carriers and agencies can work inside the general assistants they already favor and receive answers grounded in insurance specifics.
Larger enterprises go further, selecting several MCP tools, pairing them with their own, and assembling workflows the supplier may not have anticipated. Zywave is also bringing its own products closer together, for example, by moving historical loss analytics into the quote flow so that coverage recommendations reflect how comparable losses have played out. Software is turning into a set of building blocks arranged around each customer’s strategy, and as customers reach Zywave’s data through the tool of their choice, the interface matters less, and the data behind it matters more.
Spending Is Outrunning Proof
Insurers are making substantial investments in technology. The HTEC AI Value Realization Report 2026, based on a survey of 1,500 C-level executives, found that nearly 88% of financial services and insurance (FSI) respondents raised their AI budgets over the past year, by an average of 35.9%, among the steepest rises of any sector.
Looking at insurers alone, the pace differs by market. German insurers raised spending by 40.36% on average, and about 53% of US and UAE insurers increased theirs by somewhere between 41% and 50%. In the UK, 92% of insurers raised spending, well above the 79% average across all respondents, yet their mean increase of 29.98% was the lowest among the markets studied.
Spending only pays off when someone can show what it produced, which requires a traceable path from an action to its outcome, such as an agent’s outreach leading to a booked meeting and then a quote. Measurement practice is still catching up. Only a third of FSI organizations (33.1%) require a formal business case before approving an AI project, and 24.2% work out the return after the system is already live.
The most common obstacles FSI leaders cite to measuring AI value are benefits that are hard to quantify (29.2%) and analytics tools that fall short (28.8%).
Risk is handled just as unevenly, with 11.9% measuring it across their entire AI portfolio and 45.3% checking only the highest risk uses.
Doug Marquis’s advice closes the gap between spending and proof by moving measurement to the start: log how much time and money a task takes before an agent handles it, then measure the same task afterward. That gives a board a documented saving it can check for itself. Zywave itself also began with prospecting and outreach, where a poorly targeted message does little harm, instead of the work of recommending what protection a client should buy, where a mistake could violate a regulation or leave that client underinsured.
Trust Sets the Pace of Adoption
How quickly AI takes hold inside a company depends less on the technology than on the people expected to use it.
Nearly 78% of FSI leaders say people and processes limit returns more than the tools do, led by resistance to change (30.5%) and fragmented platforms (28.8%).
The companies that succeed start with tasks people feel comfortable delegating, keep a person reviewing the output, and ensure support teams can walk users through how an answer was produced.
What buyers ask of suppliers
Buyers expect the same transparency from their technology partners. FSI leaders most often request full traceability across prompts, lineage, and audit trails (28.5%), and 60% of UAE insurers rank security and privacy controls first among the areas where suppliers owe them more visibility. Explaining how an agent reached its answer has become a purchasing criterion, in addition to a support task.
What Leaders Should Do Next
When models are interchangeable, the lasting advantage lies in how well an insurer organizes and measures what it already knows. Three habits from the conversation apply to any insurer planning its next AI investment: keep the freedom to switch models, suppliers, and data sources as the market shifts, record what key workflows cost today so future savings can be proven, and bring scattered information together early, since everything built afterward draws on it.
Nearly a third of FSI executives (30.1%) say they want outside help with data foundation and governance, and HTEC supports all three habits. Its forward-deployed engineers work inside the customer’s environment to build open architecture, so switching models, suppliers, and data sources stays straightforward as the market shifts. The same team provides end-to-end visibility into adoption, usage costs, and realized value, which gives insurers the documented baseline they need to prove savings.
HTEC also organizes the scattered information that agents rely on into a single foundation they can all draw on. That experience comes from working with companies such as Zywave, and from helping clients manage the economics of AI consumption and build AI capabilities into the products and services they already sell. To hear Doug Marquis describe the choices behind Zywave’s AI strategy, watch the whole episode here:
