P&C Consumer AI Support Doubles But a Trust Ceiling Persists

From analyzing six consecutive consumer sentiment surveys on insurance AI over three years, the pattern is clear: acceptance for tools runs far ahead of acceptance for decisions, and that gap narrows at roughly five percentage points per year. Insurity’s April 2026 release of its second annual AI in Insurance Consumer Survey puts hard numbers behind that pattern. Across more than 1,000 randomly selected U.S. adults surveyed in February 2026, support for insurer AI use jumped from 20% to 39% in a single year. Purchase reluctance dropped from 44% to 36%. Yet the survey also maps a hard trust ceiling: only 16% of respondents would let AI cancel or renew a policy, and nearly half distrust AI for claim approvals, fraud determinations, or policy adjustments. For carriers planning AI deployment roadmaps, for actuaries modeling expense ratio improvements from workflow automation, and for regulators calibrating oversight thresholds, the Insurity data draws a bright line between what consumers will accept now and what they will not.

The headline numbers: 20% to 39% in twelve months

The topline finding is striking. When Insurity ran the same survey in early 2025, just one in five respondents said it was a good idea for their insurance company to use AI to improve services. One year later, that figure nearly doubled to 39%. On the resistance side, the share of consumers who said they would be less likely to purchase a policy from an insurer that publicly uses AI fell from 44% to 36%. Both moves are large for a single twelve-month period in an industry that typically measures sentiment shifts in low single digits.

The backdrop makes the shift legible. According to the same survey, 84% of U.S. consumers now use AI tools at least occasionally, with 27% reporting daily use. The baseline has moved. AI is no longer an unfamiliar concept that insurers must explain to policyholders; it is something policyholders already use for writing, workplace productivity, health queries, and financial comparisons. That normalization changes the framing. The question consumers answer in the Insurity survey is no longer “should my insurer use this unfamiliar technology?” but “should my insurer use this technology I already use elsewhere?” The jump from 20% to 39% makes more sense in that context.

The competitive penalty for AI disclosure is shrinking in parallel. At 36%, purchase reluctance remains material; more than a third of consumers still prefer an insurer that does not advertise AI. But the eight-point drop from 44% means that insurers publishing their AI capabilities on marketing materials or product pages face a smaller headwind than they did a year ago. Carriers like Travelers, which deployed Anthropic-powered AI assistants to 10,000 staff, and AIG, which has publicized its Palantir-powered agentic underwriting stack, are operating in a consumer environment that is measurably more receptive than the one they launched into.

The tool-versus-decision trust gap

The more consequential finding sits below the topline. Insurity asked respondents to rate their comfort with AI performing specific P&C insurance tasks. The results split cleanly into two tiers.

High-comfort tier (AI as assistant). 46% of respondents said they would let AI generate an insurance quote. 39% were comfortable with AI tracking claim status. 38% would allow AI to update their personal information on file. These are tasks where AI gathers, organizes, or presents information but does not make a consequential decision. The consumer retains the final action: accepting the quote, filing the claim, confirming the update.

Low-comfort tier (AI as decision-maker). Only 22% would let AI file a claim on their behalf. Just 16% are comfortable with AI canceling or renewing a policy. Nearly 50% of respondents specifically expressed distrust of AI for claim approvals, fraud flags, or policy adjustments. These are tasks where AI exercises judgment that directly affects the consumer’s coverage, financial position, or claim outcome.

The gap between 46% (quoting) and 16% (policy lifecycle decisions) is 30 percentage points. That is not a gradual gradient; it is a step function. Consumers have drawn a line between AI that helps them and AI that decides for them.

A third data point completes the picture. Only 33% of respondents trust AI-driven insurance decisions overall, with 26% saying they need more information before forming an opinion. That 26% undecided segment is where carrier communication strategy matters most. These are consumers whose trust is contingent on transparency: they will convert to either side based on how clearly carriers explain what AI does and does not do in their workflows.

Mapping the trust gap onto carrier deployment strategy

From tracking carrier AI deployments across the major P&C writers over the past three years, we see the Insurity data aligning precisely with where carriers have already placed their bets, and where they have held back.

Green-light zone: quoting and FNOL. The 46% comfort level for AI-generated quotes validates the deployment pattern already visible at scale. AIG Assist reported a 30% quoting lift and 55% time-to-quote reduction in Q1 2026 across Lexington middle-market property. Progressive’s HomeQuote Explorer uses AI to generate real-time quotes from property data. Travelers’ AI assistants support underwriting quote workflows across 10,000 staff. First notice of loss is similarly positioned: 39% comfort with claim tracking and 38% with data updates map naturally onto AI-powered FNOL intake, which collects information, routes claims, and confirms details without adjudicating outcomes.

Caution zone: claims adjudication and fraud. The sub-25% comfort levels for AI filing claims and the nearly 50% distrust of AI claim approvals and fraud determinations explain why carriers have been careful to maintain human-in-the-loop architectures for these functions. AIG’s Q1 2026 disclosure of an 88% claims AI alignment benchmark explicitly references human validation. Sedgwick’s data showing 82% carrier AI adoption but only 7% reaching full scale reflects, in part, the hesitation to push autonomous claims decisions past internal compliance thresholds when consumer sentiment sits this low.

Red zone: policy lifecycle decisions. At 16%, consumer comfort with AI canceling or renewing policies is the lowest reading in the survey. No major carrier we track has publicly deployed fully autonomous AI for policy cancellation or non-renewal decisions. The regulatory overlay reinforces this: the NAIC Model Bulletin on AI, now adopted in over half of all states, requires that insurers remain responsible for decisions affecting consumers regardless of how much AI sits in the workflow. The 16% figure provides carriers with a quantified consumer-sentiment basis for maintaining human authority over policy lifecycle decisions, independent of regulatory requirements.

Expense ratio implications: which automation is attainable now

The practical value of the Insurity data for actuarial work lies in its ability to segment workflow automation opportunities by consumer acceptance rather than by technical feasibility alone. Morgan Stanley’s projection that AI could reduce P&C expense ratios by 200 basis points and generate $9.3 billion in operating income by 2030 assumes a deployment trajectory. The Insurity trust data constrains that trajectory by identifying which automations can scale without consumer friction and which face a years-long trust-building process before they become viable.

Near-term expense ratio capture (2026 to 2027). Quoting automation, FNOL intake, claim status communication, policy data updates, and document processing fall in the high-comfort tier. These workflows carry significant labor costs. Quoting alone involves underwriting support staff, data entry, rating engine interaction, and broker communication. AI-assisted quoting that reduces time-to-quote by 40% to 55% (the range reported by AIG and Ki Insurance) compresses the headcount required per $1 million of new written premium. Similarly, automated FNOL reduces the call center and adjuster intake burden. An actuary modeling near-term expense ratio improvement can weight these workflows at their current consumer acceptance levels: 38% to 46% of policyholders will not resist, and the 36% purchase-reluctance figure suggests the marketing downside is manageable.

Medium-term opportunity (2028 to 2029). Claims adjudication automation, fraud scoring with reduced human review, and AI-assisted settlement offers occupy the caution zone. At 22% comfort for claim filing and near-50% distrust for claim approvals, carriers pushing autonomous claims decisions face both consumer backlash and regulatory scrutiny. The Insurity year-over-year trend suggests this zone will open gradually. If the five-point-per-year narrowing pattern holds for the tool-versus-decision gap, consumer comfort with AI claims adjudication may reach the mid-30s by 2028, still below majority acceptance but potentially sufficient for carriers to deploy AI-primary claims workflows with robust human escalation protocols.

Long-term constraint (2030 and beyond). Policy cancellation, non-renewal, and coverage modification decisions sit at 16% acceptance. Even at five points of annual improvement, majority consumer acceptance for AI-driven policy lifecycle decisions would not arrive until the early 2030s. For expense ratio models projecting beyond 2028, this is a hard constraint: the labor savings from automating underwriting judgment on renewal and non-renewal decisions remain largely inaccessible for the planning horizon that most carriers use.

The 26% persuadable segment and transparency ROI

The 26% of respondents who said they need more information before trusting AI-driven insurance decisions represent the swing vote in this survey. They are not opposed to AI; they are waiting for evidence.

For carriers, this segment translates into a measurable return on transparency investment. The Insurity data implies that a carrier moving from generic AI disclosure (“we use AI to improve your experience”) to specific workflow disclosure (“AI generates your quote based on property data and claims history; a licensed underwriter reviews it before binding”) can shift a meaningful share of the undecided 26% into the acceptance column.

Deloitte’s 2026 Global Insurance Outlook reinforces this finding. While 90% of insurance executives agree on the urgency of redesigning the employee value proposition for human-machine collaboration, only 25% have taken tangible action. The 65-point intent-to-action gap that Deloitte, Oliver Wyman, and McKinsey all identified extends to consumer communication: carriers know they need to explain AI to policyholders, but most have not operationalized that explanation.

J.D. Power’s 2026 U.S. Property Claims Satisfaction Study provides supporting evidence. Digital tool adoption drove meaningful satisfaction gains: 38% of claimants used digital FNOL, 49% submitted photos digitally, and 45% received digital updates. Overall satisfaction rose 20 points to 702 on a 1,000-point scale. Consumers who used digital tools at each claims touchpoint reported higher satisfaction than those who did not. The pattern is consistent: when consumers can see what the technology does and retain control over the outcome, satisfaction rises. When the technology operates opaquely or autonomously, trust erodes.

Accenture’s 2026 insurance executive survey found that 90% of global insurers plan to increase AI spending this year. EY reported that 69% of insurers are focusing on agentic AI for risk assessment and mitigation, while 60% prioritize customer service and engagement. The supply-side investment is accelerating. The Insurity data tells carriers where that investment can land without running into a consumer trust wall, and where the wall still stands.

Regulatory alignment: the NAIC overlay

The Insurity trust data and the NAIC regulatory posture are converging on the same answer from different directions. The NAIC Model Bulletin, adopted in over half of all states, requires that insurers maintain accountability for decisions regardless of AI involvement. The 12-state AI Systems Evaluation Tool pilot running from March to September 2026 gives examiners a structured framework for reviewing AI governance during market conduct examinations. The four-tier AI risk taxonomy proposed at the Spring 2026 National Meeting classifies AI applications by risk level, with high-risk designations for applications that affect consumer access to coverage.

The alignment is notable. Regulators categorize policy cancellation and claims adjudication as high-risk AI applications. Consumers, independently, put those same functions at the bottom of their comfort rankings. This convergence creates a double constraint on carrier deployment: even if technical capability permitted fully autonomous claims adjudication or policy lifecycle decisions tomorrow, both consumer sentiment and regulatory posture block it.

For actuaries involved in rate filing support, this matters directly. State rate filings that incorporate AI-generated pricing models must demonstrate compliance with the NAIC framework. The Insurity data provides external validation for the human-oversight requirements that regulators are embedding in examination protocols. An actuary defending an AI-augmented rate filing can point to the consumer sentiment data as independent evidence that the human-in-the-loop design reflects policyholder expectations, not just regulatory mandates.

Competitive positioning: the disclosure dilemma

The eight-point decline in purchase reluctance (44% to 36%) creates an evolving competitive calculus. A year ago, disclosing AI use was a net negative for customer acquisition: more consumers were deterred than attracted. The 2026 numbers are closer to neutral. At 39% support versus 36% reluctance, the spread has flipped from negative-24 to positive-3. Carriers that disclose AI use may now attract marginally more policyholders than they lose.

That said, the margin is thin and the aggregate hides variation. The Insurity survey was conducted across 1,000+ randomly selected adults, and the 18-question instrument covered a range of insurance AI scenarios. Demographic variation by age, income, geography, and prior claims experience likely drives significant within-sample dispersion. A 39% average support figure could mask 55% support among younger, digitally native consumers and 25% support among older policyholders with longer carrier relationships. Carriers serving different demographic segments will face different disclosure equations.

The competitive implications extend to the carrier AI strategies we have covered in this series. Chubb disclosed plans for a 20% headcount reduction through AI-driven automation, projecting 1.5 combined ratio points in expense savings. Progressive has invested heavily in AI-powered telematics and quoting. AIG has made its Palantir-powered agentic underwriting stack a public differentiator. Each of these carriers is making a bet that AI disclosure, whether framed as efficiency, accuracy, or speed, will become a competitive asset rather than a liability. The Insurity trend line supports that bet directionally, but the 36% residual reluctance means the transition is not complete.

What the trend line predicts

Two years of Insurity survey data are not sufficient for a robust time-series forecast, but combined with other consumer sentiment benchmarks, the direction and approximate velocity are legible.

Consumer support for insurer AI use rose from 20% to 39% in one year, a 19-point jump. Purchase reluctance dropped eight points. Comfort with AI quoting sits at 46%, and comfort with AI policy decisions sits at 16%. If the tool-tier comfort levels continue growing at five to eight points annually, while the decision-tier comfort grows at three to five points, the two tiers will converge slowly.

By 2028, based on the current trajectory, AI quoting comfort may reach 55% to 60%, making it a majority-accepted capability. AI claims involvement may reach 30% to 35%, still below majority but approaching the threshold where carriers can deploy AI-primary claims workflows without facing significant consumer friction. AI policy lifecycle decisions may reach 25% to 30%, still a minority position and still a binding constraint on automation.

The gap narrows but does not close within any reasonable planning horizon. For the next five years, the assistance-versus-autonomy divide will continue to shape which AI investments deliver expense ratio improvements and which remain technically ready but consumer-blocked.

Actuarial implications: four takeaways

1. Expense ratio models should incorporate consumer acceptance as a deployment constraint. Technical AI capability and consumer willingness to interact with AI are not the same input. Morgan Stanley’s 200-basis-point expense ratio projection assumes a deployment ramp. The Insurity data quantifies where that ramp hits friction. Pricing actuaries building multi-year expense ratio assumptions should discount automation savings for workflows below 30% consumer acceptance and weight savings at full value only for workflows above 40%.

2. The quoting and FNOL expense reductions are real and near-term. At 46% and 39% consumer comfort respectively, AI-assisted quoting and AI-powered FNOL intake face minimal consumer friction. The labor savings from these workflows, including underwriting support, data entry, call center staffing, and adjuster intake, are the most accessible expense ratio improvements in the current environment. Carriers that have already deployed in these areas (AIG, Progressive, Travelers, Ki Insurance) are capturing these savings now.

3. Claims automation savings require a longer runway. At 22% comfort for AI claim filing and near-50% distrust for claim approvals, actuaries should model claims automation benefits with a delayed onset. The savings are real in technical terms. Deloitte projects $80B to $160B in fraud savings from AI, and carrier earnings calls increasingly reference AI claims efficiency. But the consumer trust constraint means that full-scale deployment of autonomous claims AI, the kind that would materially reduce loss adjustment expense ratios, is several years away for most carriers.

4. Human-in-the-loop design is a pricing assumption, not just a compliance feature. The regulatory overlay (NAIC Model Bulletin, 12-state evaluation pilot, proposed four-tier risk taxonomy) and the consumer sentiment data point in the same direction: human oversight of consequential decisions will be required for the foreseeable future. That means the labor cost of human reviewers, escalation handlers, and decision auditors remains a permanent component of expense ratio assumptions for claims adjudication and policy lifecycle functions. Actuaries should not model these costs out of run-rate assumptions within a five-year horizon.

Why this matters

The Insurity survey is the first year-over-year benchmark on consumer AI acceptance specific to P&C insurance. It confirms what carrier deployment patterns have implied: the industry has a green light for AI tools and a red light for AI decisions. The 19-point jump in overall support is encouraging for carriers investing heavily in AI infrastructure. The 30-point gap between quoting comfort and policy-decision comfort is the number that constrains deployment strategy.

For the actuarial profession specifically, the data resolves an assumption that has been implicit in most carrier AI business cases: the assumption that consumer acceptance will not be a binding constraint. It is a binding constraint, and it has a measurable shape. Quoting, FNOL, and data management are open. Claims adjudication is opening. Policy lifecycle decisions are closed. Expense ratio projections, deployment timelines, and staffing models should reflect that segmentation.

The 26% undecided segment is the variable that carriers can influence. Transparent disclosure of what AI does in each workflow, combined with clear communication about where human judgment remains, is the lever that moves consumers from undecided to accepting. The carriers that invest in that communication will capture the automation dividend sooner. The ones that deploy AI without explaining it will run into the trust ceiling.