Carriers succeed with AI by fixing culture and processes first

To find the most success with AI, insurance companies should view it more as a way to fundamentally rework processes — and even their corporate culture — than as a mere tech deployment.

This is a recurring theme among the industry experts who write for Digital Insurance's InsureThink.

"There is a natural temptation to begin an AI initiative by asking what the technology can do and where you can place it within your business," writes Amy Carlisle, president of MSI. "I have found it more useful to start with understanding your core processes and friction points."

Tatum Fish, a strategic value architect at Cytora, framed the same issue as one of company culture.

"I've seen the best results come from teams where people comfortable experimenting with new tools work alongside those who understand where the real complexity and risk sits," Fish wrote. "Ultimately, digital transformation isn't measured by how many AI tools an insurer deploys. It's measured by whether those tools genuinely improve the way people work."

Read more from our InsureThink thought leaders:

AI speeds quoting, but accountability must stay human

Insurers rushing to deploy AI should target processes that create bottlenecks — such as policy intake, claims documentation or renewal prep — rather than starting with the technology itself, writes Amy Carlisle, president of MSI. At MSI, AI cut quoting time for complex real estate portfolios from four to five days to under 24 hours, and reduced high-net-worth home risk reviews from 20-30 minutes to a few minutes per file. Carlisle stresses that accountability for final decisions must remain with underwriters and process owners, not automated tools. Firms should track how freed-up time improves judgment, broker response and product development, while maintaining ongoing oversight of AI accuracy and compliance.
Read more: Why losing accountability is the real AI risk

Build explainable AI trails to defend fraud decisions

A Verisk study found that 98% of insurers say AI editing tools are driving digital fraud, but only 32% feel confident spotting deepfakes. Aviva flagged $311 million in suspected claims fraud in 2025, citing AI-manipulated images and documents. Rather than relying on black-box AI fraud scores that leave adjusters without clear evidence, claims teams should adopt tools that surface explainable data points — linking images, records, timelines and claimant behavior — to build a defensible paper trail. Tom Rasmussen, VP of product for claims at Carpe, argues that AI should augment, not replace, adjuster judgment, ensuring fraud decisions withstand legal scrutiny and limit reputational risk.
Read more: Why AI fraud tools could cause claim risks

Pair AI adoption with process redesign, insurers urged

Digital transformation stalls when insurers bolt AI onto broken workflows instead of redesigning processes, writes Tatum Fish, strategic value architect at Cytora. She urges underwriting leaders to start technology investments with a specific operational problem — slow submissions, inconsistent manual work or broker delays — rather than adopting AI for its own sake. Carriers seeing the strongest returns pair new tools with process redesign and blend experienced underwriters' judgment with younger employees' comfort challenging legacy workflows. Fish argues that this generational "productive friction" pushes carriers to question outdated practices while preserving critical underwriting expertise, making technology and culture change inseparable drivers of long-term advantage.
Read more: Why insurers need technology and culture to evolve together

Audit integration before buying new agency tech

Agencies often blame front-end platforms for operational bottlenecks when the real culprit is poor integration between core systems, writes Crystal Darling, senior director of retail tech services at ReSource Pro. A targeted API investment — sometimes just a few hundred thousand dollars — can resolve friction that a multimillion-dollar platform replacement won't fix. Before signing new contracts, mapping current-state workflows to pinpoint actual friction points is essential, along with demanding transparent APIs and accessible data from vendors and requiring pilot programs before enterprise-wide rollouts. Clean, governed data remains foundational to supporting automation, AI and analytics without fragmenting agency operations further.
Read more: Poor integration can tank agency tech rollouts

90% of AI agent risk hides in silent coverage

More than 90% of insurers' AI agent exposure sits as silent coverage inside cyber, D&O, general liability and tech E&O policies never built for autonomous systems, per the Underwriting the Agent Economy report. Insurers should mandate that policyholders preserve version-specific logs, prompts and permission records at deployment, since retention rules erase evidence needed after a loss. Discounting premiums for verifiable AI governance — as nuclear insurers cut rates up to 40% and cyber insurers up to 25% for strong security evidence — could reward better recordkeeping. Dr. Viroshan Naicker, CEO of Refiant, notes that only one in five firms has mature governance for autonomous agents, leaving underwriters overconfident in risks they can't fully trace.
Read more: How insurers might cover risks AI agents create

Commission entry eats 4,000 hours a month, AI can fix it

Agency finance teams spend more than 4,000 hours a month on commission entry and reconciliation, according to industry benchmarks cited by Nash Qadri, vice president of product management and development at Applied Systems. As AI adoption in insurance jumped from 8% to 34% year-over-year, agencies can automate commission matching, embed premium financing into core systems and shift month-end close from a weeklong scramble to a continuous process. Before adopting AI-driven finance tools, agencies need to clean up inconsistent policy coding and outdated carrier data, and press vendors on who controls financing relationships if a platform is acquired or changes terms — critical to avoiding costly vendor lock-in.
Read more: Can AI turn agency finance from cost center to strategist?

Chemical incidents reveal gaps in pollution liability plans

Recent chemical tank failures in Washington and California — one fatal, one triggering mass evacuations — have renewed scrutiny of pollution liability coverage. Many general liability policies fully exclude pollution risks, leaving insureds unaware of gaps until a claim hits. Blended GL-pollution policies help but share aggregate limits, so one large pollution claim can drain coverage for both exposures. Dennis Willette, SVP and Head of Westfield Specialty Environmental at Westfield Specialty Insurance, points to standalone pollution policies as the strongest option, offering dedicated limits and broader terms. Brokers should use these incidents to prompt clients to reassess risk tolerance and coverage adequacy now, before the next claim exposes the gap.
Read more: Why brokers and insureds need a new pollution liability strategy

Modern MMM lifts insurer marketing ROI

Insurers relying on last-click attribution risk misallocating marketing budgets, with legacy models sometimes overstating channel contribution by more than 40%, says Karl Canty, partner at Capco. As multi-touch tracking degrades under tightening privacy rules (NAIC, CCPA, NYDFS), carriers are turning to AI-enabled marketing mix modeling (MMM), which uses aggregate, privacy-safe data instead of individual-level tracking. Early adopters report 10% to 30% gains in marketing ROI. Building unified data foundations linking spend to policy outcomes, adopting Bayesian models with confidence ranges, and separating acquisition from retention spend can make budget decisions — and CFO scrutiny — more defensible and audit-ready.
Read more: How insurers use AI with MMM to fix wasted marketing spend

36% of small-business owners now turn to AI for coverage

Small-business owners increasingly research coverage through AI tools before contacting an agent, with 36% using AI for insurance research versus 31% consulting an agent, according to ERGO NEXT Insurance survey data. Publishing clear, structured content — FAQs, coverage guides, thought leadership — across channels where AI and search tools pull information helps insurers surface accurate, discoverable expertise before prospects reach a website. Streamlining digital experiences with personalization, progressive data collection and conversational AI reduces friction once customers arrive, while preserving clear pathways to human advisors remains critical, since one in five uninsured owners cite confusion as a barrier, notes Effi Fuks-Leichtag, chief product officer at ERGO NEXT Insurance.
Read more: What to do when small businesses trust AI as much as agents

Vendor incidents average $145K per claim, data shows

Vendor-related incidents accounted for 14% of cyber claims in 2025 and averaged $145,000 per claim, according to Deborah Dioguardi, EVP and professional lines national practice leader at Jencap. Brokers should confirm clients vet every vendor's security posture, require proof of vendor cyber liability insurance and build contractual protections before an incident occurs — not after. Review policies for dependent business interruption coverage, and flag sublimits or exclusions tied to supply chain events or improperly vetted vendors that can quietly erode protection. Assessing concentration risk, or how much revenue or data hinges on a single vendor, exposes gaps most policies still miss.
Read more: What happens when your client's vendor gets hacked?

Property-level modeling beats ZIP code risk assessment

Neural network-based embeddings can outperform ZIP code and territory-level segmentation by capturing property-specific risk factors — construction, maintenance, local loss history and evolving weather exposure — that traditional models miss, according to Brian Keller, vice president of AI and data science at LexisNexis Risk Solutions. Insurers adopting this granular approach can sharpen underwriting by flagging hidden risk in attractive areas and identifying safer properties in challenging regions, price policies to match actual exposure and reduce cross-subsidization, and track risk shifts across the policy lifecycle for earlier intervention. Portfolio managers also gain better tools to control concentration, curb adverse selection and target growth in underpriced markets, shifting risk assessment from geographic averages to the individual property.
Read more: Neural networks spot risk differences that ZIP codes miss

Build an evidence spine to meet EU AI Act demands

The EU AI Act applies based on market use, not company location, meaning U.S. insurers can fall under its scope through European customers, partners or deployments. Compliance will often surface first in procurement questionnaires, not regulatory audits. Rather than building separate compliance systems per jurisdiction, insurers should create a reusable "evidence spine" — a unified record linking AI outputs to source documents, model versions, user access and human review. This structure should integrate with existing claims and underwriting systems rather than sit as a separate AI layer. Corto Romagny, General Manager, U.S. Operations at Uxopian Software, argues that this approach turns governance into a competitive, market-access advantage.
Read more: U.S. insurers will need an evidence spine for the EU AI Act

This roundup was created with AI assistance. A Digital Insurance editor reviewed each item before publication.