Chubb made headlines in December 2025 when it announced a 20% headcount reduction tied to AI automation.

Around the same time, the CEO of a regional mutual carrier in Indiana made the opposite bet. Wes Sprinkle, president and CEO of Indiana Farmers Insurance (IFM), decided that AI would elevate rather than replace his team. The company would not pursue AI-related layoffs, but AI proficiency would become a requirement for its 44-member leadership team first, and later, all 250 employees.

But integrating AI into decades-old workflows at a 150-year-old mutual insurance provider required the kind of cultural change management that most carriers have never had to undergo. AI transformation is a culture problem dressed up as a technology and training problem, and we knew we had to treat it as such.

The Lippitt-Knoster Model for organizational change management shows successful change requires five elements: vision, skills, incentives, resources, and an action plan. Miss any one of these and you get predictable failure modes.

• No vision? Confusion.
• No skills? Anxiety.
• No incentives? Resistance.
• No resources? Frustration.
• No action plan? False starts.

IFM addressed all five with Pragmatico’s help. The results quickly followed.

In just 16 weeks, daily AI usage among leadership climbed from 47% to 91%. Confidence in using AI jumped from 19% to 77%, a more than 3x increase. The cohort was saving more than 218 hours per week, which translates to roughly $1.05 million in productivity gains with zero headcount reductions.

Mapping the five Lippitt-Knoster elements to an insurance organization required a sequence. Vision and skills had to come before incentives. Incentives had to come before sustained behavior change.

The work fell into three phases, driven by IFM’s senior leadership team with Pragmatico as the facilitator:

Step 1: Align on AI Vision, Policy and Baseline Metrics

Most AI initiatives launch with tool license rollouts, a kickoff email and a smattering of training sessions, skipping the architecture work that makes the rest of it stick. Three foundational pieces need to be in place before anyone touches the tools.

The first is a leadership-authored AI vision. This is a short document, ideally one page, that outlines the organization’s beliefs about AI and how it will be used. At IFM, the document went through three revisions. The final version connected AI adoption directly to IFM’s ICARE values (Innovation, Collaboration, Accountability, Respect, Empowerment). It committed to integrating AI proficiency into performance management, job descriptions and hiring criteria within 6-12 months. And it included a line that Sprinkle insisted on after reading about Chubb’s AI-driven headcount reductions: AI would be about elevating people, not replacing them.

Second, we thoughtfully crafted an AI policy. This is the workstream where a lot of insurers get stuck. Compliance teams often write conservative policies that prohibit personally identifiable information from being entered into any AI tool. In practice, that means claims adjusters cannot use AI for the exact workflow where it would deliver the most value. The right approach is a tiered framework: public, internal, confidential and restricted categories, each with a clear yes-or-no on enterprise AI use, calibrated to the data protections the enterprise platform actually provides. (In other words, sort information into four sensitivity levels and clearly state whether the company’s AI tools may be used with each one, based on the actual safeguards those tools have.)

Third, we deployed a baseline survey before any training began. Done well, the survey establishes a true starting point for each individual and the cohort, which makes progress measurable. It also surfaces the emotional landscape of the cohort: who is enthusiastic vs. anxious, who has never touched the technology, and what specific concerns need to be addressed. And the act of deploying it, with CEO sponsorship and a clear connection to the broader initiative, signals to the organization that AI is a serious commitment rather than optional enrichment.

Step 2: Activate Behaviors With Training and Built-in Accountability and Recognition

AI training fails in predictable ways. A vendor delivers a 60-minute webinar. Employees nod along blankly, play with the tools once and go back to the way they’ve always worked. At IFM, training was delivered across three levels over 90 days, with each session building on the previous one.

Accountability was baked in from the beginning. Homework on specific applications was assigned after each session. A usage leaderboard, drawn from the enterprise platform’s admin analytics, was displayed at each training session. AI “wins” were celebrated in a dedicated Teams channel: a claims supervisor reducing a one-hour liability review to 10 minutes, a communications leader loading 18 months of agent emails into a reusable system, and an IT lead processing 800,000 log entries in a single query.

CEO Sprinkle stayed visible throughout, attending training and using the tools himself. Active and visible sponsorship is the single biggest predictor of change initiative success according to Prosci research, and it cannot be delegated to a committee or to IT.

Step 3: Operationalize and Sustain Behavior Change Long-Term

The fastest way to lose the gains of a time-bound program is to declare victory and move on. Sustained behavior change requires measurement and a clearly communicated set of expectations around AI proficiency.

The metrics that matter are workflow-level: how many people on the team are using AI at least weekly, where usage patterns are clustered, what cycle times are changing, what error rates and quality audits are moving. At IFM, message volume through the enterprise AI platform nearly doubled in three months, growing from 1,928 messages in November 2025 to 5,231 in February 2026. The patterns inside those numbers told the operational story. Departments where the manager engaged personally showed broader team adoption. Departments where the manager had not engaged showed lower usage. That insight drove targeted coaching.

AI proficiency also belongs in performance management. Every leader has 2026 goals around leading with AI. This is what reinforcement looks like in change management terms. New behaviors become the expected behaviors, and the organizational systems that reward and recognize people align with new expectations.

The ROI was already concrete before IFM began rolling out the program to all 250 employees: 218 hours per week saved across 44 leaders translates to roughly 10,900 hours annualized, or $1.05 million in incremental capacity value.

What This Means for the Industry

Research has found that roughly 70% of insurers are experimenting with AI tools in their operations but only 7-10% have achieved scalable AI success. Nearly two-thirds of carriers acknowledge a disconnect between their AI vision and current reality, even as industry AI investment is projected to soar.

Carriers that treat AI adoption as cultural transformation will see the gap between their AI vision and their AI reality close quickly. The carriers that do not will continue funding pilots that stall, while their best underwriters and claims professionals quietly use whatever tools they can find, with or without permission.

Buying the tools is easy. Building a culture that actually uses it is much harder. But the latter is what will actually drive transformation.

Disclosure on AI use: The authors used AI tools to assist with research synthesis and structural editing of internal source material. All facts, data points and conclusions were reviewed and verified by the authors, who remain accountable for the accuracy and originality of the content.

Top photo: AI-generated image: ChatGPT