The insurance market appears ready to shift repeatable operational work toward AI. The goal is broader capacity, including business insurers currently leave unquoted.
According to ISG research commissioned by mea Platform, 83% support AI execution. The figure covers repeatable work across major insurance operating functions.
Trust comes with clear conditions across much of the market. Some 75% want insurance-specific models or systems governed by internal rules. General-purpose AI alone attracts far less confidence for consequential decisions.
ISG surveyed senior executives working across several insurance functions. Respondents represented underwriting, operations, claims, technology and transformation leadership. The research examined 20 operational activities across major insurance markets. Participants came from North America, Europe and Asia.
Those activities include submission intake, triage and quote generation. Bordereaux processing also appears within the study’s operating scope. Claims adjudication and compliance screening form part of the research as well.
Human judgment still carries greater weight for consequential business decisions. According to the study, 86% want those decisions kept with people.
Respondents instead favor AI handling repetitive work beneath those decisions. The technology would prepare evidence before human judgment enters the process. Underwriters and brokers then reach risk decisions with stronger supporting information. The intended result is also faster decision-making.
Operational capacity already affects insurers’ ability to quote desired business. Carriers estimate one in nine broker submissions never receives a quote.
Some submissions are declined even when appetite exists for the risk. Others remain unquoted because operating teams cannot process them quickly enough.
This capacity constraint puts operational performance directly against potential premium growth. Insurers aren’t only looking for lower administrative expense through AI deployment.
Among insurers already using AI operationally, 61% report productivity gains. Another 51% report faster cycle times after deployment.
Respondents also expect operating costs to decline 16% over two years. Cost remains part of the business case, though capacity carries wider commercial value.
The broker relationship adds another measure of operational performance. Asked what would improve broker perceptions most, 64% selected pricing.
Ease of doing business followed at 52% among respondents. AI-driven speed and submission completeness received 51%. Those results place operational execution close to established commercial considerations. Faster processing gives carriers more opportunity to respond before business moves elsewhere.
Insurers draw a clear boundary around high-consequence underwriting and claims decisions. Limited human oversight raises the threshold further. For those decisions, 75% selected insurance-specific models or governed hybrid systems. Only 6% accepted a general-purpose model operating by itself.
The study gives governance a defined operational meaning. AI systems work from controlled policy wording and approved endorsements. They also follow underwriting appetite and established claims guidance. Permissions separate information access from authority to make decisions.
A designated person retains authority to override or stop the system. The institution also sets appetite limits and referral triggers.
Thresholds remain tied to the insurer’s existing operating rules. The model therefore executes established practice instead of creating new standards independently.
Martin Henley, Chief Executive of mea Platform, linked those controls with growth. He said insurers are prepared to transfer repeated work to AI. Their condition is an insurance-aware model operating within institutional rules. Henley described that requirement as demanding but appropriate.
He also pointed to business insurers already want to write. Operational bottlenecks currently prevent some of those opportunities reaching a decision.
For carriers, recovering that business carries greater value than cost savings alone. AI therefore becomes a capacity issue as much as an efficiency project.
AI-led operating redesign already appears on most insurers’ agendas. The study puts that proportion at 96%. Current deployment remains far behind those stated plans. Only 13% report an advanced operating posture today.
Under that structure, AI sits centrally within workflow design. Respondents expect the figure to reach 52% within two years.
Fully AI-native operations remain rare across the surveyed market. Fewer than 1% currently report operating under that model.
The study defines AI-native operations around end-to-end process execution. AI handles defined processes while people establish policy and manage exceptions.
Within two years, 12% expect to reach that operating position. The gap shows how much implementation work still sits ahead.
Ashish Jhajharia, Insurance SME and Principal Analyst at ISG, addressed this difference. He said insurers understand where human judgment should remain.
The unresolved issue is how companies reach the desired operating structure. AI-led redesign sits on the agenda for 96% of respondents.
Only 13% have already reached an AI-centered operating model. Jhajharia said this difference will shape work over the next two years.
Moving toward AI-native insurance requires changes beyond model deployment. Decision rights need explicit definition before systems execute work independently.
Referral rules also determine when human involvement becomes mandatory. Firms need an evidence trail showing why systems acted as they did.
Those controls let AI act within an institution’s established authority. They also preserve human intervention when a case exceeds approved thresholds.
Insurers still place model choice inside those governance requirements. They want systems developed for insurance or constrained by internal rules.
This helps explain why 27% remain in evaluation rather than construction. Their next step depends on operating governance as much as technology selection.
