(Image source: IIR.)
When Mike Dwyer, EVP and CTO, EIS (San Francisco), spoke with IIR in October 2025, he described CoreGentic—the firm’s AI and agentic orchestration layer for its OneSuite platform—as the culmination of an architectural strategy EIS had pursued for more than a decade.
Nearly a year later, Dwyer says the most important development is the “operationalization” of that vision. EIS now applies AI to production implementation programs, has released its agentic core to customers and is introducing agents for insurance business functions.
The larger test, however, is not whether EIS can demonstrate compelling AI capabilities. It is whether those capabilities materially change the economics of implementing and operating insurance technology.
Dwyer believes insurers should expect such a dividend. AI should reduce the cost, time and labor required for core transformation, improve the economics of insurance operations and ultimately enable profitable growth. EIS must also prove that argument through its own performance.
“How do we scale significantly without adding to our workforce fundamentally?” Dwyer says. “We have to deliver that outcome ourselves internally.”
From Vision to Operationalization
Dwyer organizes EIS’s AI strategy around three pillars: applying AI to software delivery and implementation; embedding AI and agentic capabilities within the core; and applying agents to insurance business functions.
EIS has moved its AI-assisted delivery approach from pilots into production programs, according to Dwyer. The company applies AI across requirements development, configuration, integration, testing, deployment and environment configuration.
Mike Dwyer, EVP, CTO, EIS.
A core transformation remains broader than configuring the core itself, he cautions. It includes changes to the insurer’s surrounding technology ecosystem and business processes. Nevertheless, Dwyer says AI is already producing significant savings in time, cost and labor within the areas under EIS’s control.
EIS has also released its agentic core for customer use and is expanding the governance framework around it. That includes an agent software development kit through which an agent can be assessed and approved before entering production. The third pillar consists of agents supporting functions such as life insurance cash management, pricing analytics, claims processing and billing.
The maturity of these capabilities varies. AI-assisted implementation is being used in production programs, while conversational changes to production insurance processes face greater carrier hesitation over governance, approvals and operational risk.
Dwyer says many insurers have not yet fully absorbed the potential benefits of AI, much less agentic AI. For now, carriers are more comfortable applying the technology to controlled delivery processes and information-gathering tasks than allowing it to make consequential changes to production systems.
Compressing the Work
Dwyer offers product introduction and pricing as examples of how AI can alter core-system delivery.
The traditional sequence still begins with a business objective or product filing and proceeds through requirements, design, configuration and testing. AI does not eliminate those stages. It compresses the translation work between them.
CoreGentic can consume conversations and product documents, generate requirements that a configuration agent can interpret, create the corresponding configuration and produce test cases derived from the same specification. Subject-matter experts remain responsible for reviewing and approving the results.
“That journey from A to Z doesn’t change,” Dwyer explains. “It’s just that the human labor required for all those steps either goes away or becomes an approval step.”
In pricing, for example, EIS has added a conversational interface to its OpenL-based pricing engine. The system can consume a rate book, convert it into an executable pricing engine and generate test cases. An actuary can interact with the system conversationally, review the output and authorize deployment.
The significance lies less in eliminating a particular stage than in reducing the handoffs among business specialists, requirements analysts, developers, configuration specialists and testers. The process moves closer to the person who understands the insurance product, while human responsibility shifts toward judgment, validation and approval.
From System of Record to Execution
AI has prompted competing predictions about the future of core systems. One view holds that AI makes the core more important as the authoritative source of customer, policy, billing and claims information. Another suggests that agents will blur application boundaries and diminish the importance of the systems beneath them.
EIS aligns more closely with the former view. Insurance transactions still require dependable execution, security, governance, auditability and traceability, Dwyer argues. Those capabilities already reside in the core and should not have to be reconstructed in an external AI layer.
Sara Pérez, CMO, EIS.
However, Sara Pérez, CMO, EIS, resists characterizing the core as merely a system of record. AI makes it a more active participant in business operations.
“There is an evolution from system of record to a system of execution,” Pérez says. The core not only informs the business but increasingly enables it to execute and continuously adapt its operations.
That evolution does not mean every conventional interface disappears. Dwyer considers it unrealistic to expect all insurance users and processes to move quickly from graphical interfaces to conversational interaction. Insurance combines human interactions with asynchronous processes, and many employees still need systems that organize and direct their work.
A claims adjuster, for example, should not have to ask repeatedly what requires attention. The system should intelligently identify and present the appropriate tasks.
“I shouldn’t have to ask questions all the time,” Dwyer says. “I want to be guided in a certain direction. I need to know what to do next.”
AI Within the Core
EIS describes CoreGentic as intelligence embedded in the core rather than an AI layer attached to its perimeter.
Dwyer says EIS agents execute within the same platform fabric as the company’s core insurance services. They use the platform’s event infrastructure, security controls, workflows and insurance context rather than operating as a separate layer connected only through APIs.
That placement is meant to give agents governed access to insurance capabilities while preserving the controls associated with core transactions. EIS records the instructions sent to the large language model, its output, the conversational context and subsequent actions. Existing roles and privileges govern what an agent can access.
Embedding AI does not mean granting it unrestricted transactional authority. Dwyer says EIS currently permits AI to read information, interpret it and contribute to workflows, but not independently authorize actions that commit transactions.
“There is no use case where there’s any authority of action granted to AI from the write cycle,” he says. “That is human in the loop always.”
Agents can execute tasks within governed workflows, but consequential transactions require human authorization. That distinction defines the present boundary between agentic participation and autonomous action within EIS.
Dwyer also rejects the idea that embedded AI should create another form of vendor lock-in. Carriers must be able to choose their large language models and model providers, use their own agent infrastructure and integrate externally developed agents. They should not become beholden to a particular AI technology or agent framework.
“We’re not the only game in town,” Dwyer says. “We can never be the only game in town in an enterprise ecosystem.”
Can the Core Keep Pace?
MACH Alliance research argues that composable architecture enables faster AI deployment and reduces failures associated with architectural and integration limitations.
For Dwyer, the causal connection is speed of change. AI may make it possible to design a new quoting, servicing or customer experience rapidly. The benefit is constrained if the underlying core requires months to introduce the necessary product, rule or workflow.
“It doesn’t help that I can deliver a conversational interface to enable quoting for a product if I can’t introduce that product quickly,” he says.
That leads to the fundamental question Dwyer believes insurers should ask their providers: “Can the core system provider change as quickly as the agentic capability needs you to change?”
Composability contributes to that adaptability, but Dwyer cautions that it is not conclusive by itself. Separating a suite into components or exposing functionality through APIs does not necessarily produce a forward-looking platform with a lower total cost of ownership. Composability is better than monolithic architecture, he says, but does not guarantee a genuinely cloud-native or readily adaptable core.
EIS seeks MACH certification partly to distinguish verified architecture from broadly used marketing claims. Core-system providers routinely describe their technology as modern, cloud-native and composable. Certification subjects those claims to peer review against defined technical principles.
Modernization and the AI Dividend
Pérez compares the current AI cycle with the first wave of InsurTech competition. Early InsurTechs were sometimes presented as alternatives to traditional core providers, but many concentrated on colorful customer experiences layered over incumbent infrastructure.
She sees a danger that insurers will repeat that pattern with AI—adding an attractive interface without addressing the architecture and processes beneath it. AI therefore intensifies the modernization imperative.
At the same time, Dwyer argues, AI can reduce the risk of modernization. If it compresses requirements analysis, configuration, testing and deployment, it reduces the time and money exposed to a transformation program.
Dwyer describes the potential AI dividend along three vectors: delivery, operations and growth.
The delivery dividend comes from reducing the time, labor and cost required to implement a core platform and replace legacy technology. Dwyer stops short of prescribing a numerical reduction because transformation encompasses business and organizational change beyond core configuration. Nevertheless, he says carriers should demand meaningfully better economics.
“Carriers need to expect that there are significant unit-economic benefits from using AI,” he says.
EIS has little interest in preserving implementation labor for its own sake, Dwyer adds. The company benefits when an insurer’s business operates on its platform, rather than from maximizing services revenue during the transformation.
“It’s not about delivery-services revenue,” he says. “The closer that goes to zero, the better for us.”
The operating dividend comes from reducing the human effort required to support sales, service and operations. The growth dividend comes from using agentic capabilities and customer context to increase quoting efficiency, serve customers more effectively and expand revenue without proportionately increasing headcount or technology cost.
Dwyer summarizes the progression as “deliver, operate, grow.” In his view, the eventual winners will be carriers that advance to the third stage and use AI not merely to reduce expense but to create growth.
Pérez connects that potential with a changing relationship between business and technology. Among EIS customers, she says, business decisions are having a more immediate effect on technology implementation.
“We are enabling the business to dictate what the technology needs to deliver,” she says.
Beyond SI Labor Arbitrage
Reduced implementation labor also challenges the traditional SI model. Dwyer expects AI to diminish the value of large labor forces dedicated to requirements, configuration, testing and other delivery work.
“Less on the labor arbitrage for delivery, because that’s going to really go away,” he says.
That does not mean SIs cease to matter. Their value shifts toward helping insurers redesign processes, reimagine operations and convert new technology capabilities into measurable operating and growth results.
The AI dividend therefore does not consist simply of removing people from implementation projects. It shifts expertise from translation and routine configuration toward governance, validation and business reinvention.
EIS Faces the Same Test
EIS has advocated open architecture, data fluidity and cloud-native technology since well before the current AI cycle. That raises a difficult question: Why has its technical foresight not translated into visibility or market share comparable to that of larger core providers?
Dwyer says that is beginning to change, citing EIS’s position in employee benefits and increasing traction among life insurers and European carriers. Pérez says consolidation and pressure for faster product introduction increasingly favor EIS.
“Time to market is the worst enemy for the traditional players,” Pérez says. “Time to market is our best ally at the moment.”
Technology is not the only factor governing carrier selection, she adds. Choosing a less established provider requires institutional courage in an industry whose risk aversion often favors familiar vendors.
Dwyer urges carriers to conduct deep technical due diligence rather than accept labels such as AI-native, agentic, composable or knowledge-driven. The decisive test is whether the platform can adapt and scale at the speed required by the experiences AI makes possible.
EIS now faces a similar demand for evidence. Within the next year, Dwyer says, its thesis should be visible in broader adoption, improved delivery economics, better operating economics for customers and a larger EIS footprint.
It should also appear in EIS’s own ability to support more implementations without proportionately increasing its workforce. If AI is truly changing core-system economics, the result should be more than an engaging interface or agentic demonstration. It should be visible in the cost of transformation, the efficiency of operations, the capacity to grow—and the performance of the technology provider promising those benefits.