Getting to “yes” with AI adoption in insurance

Introduction 

Through Celent’s 2026 AI in North American Insurance survey, insights from Celent’s AI Bootcamps and our extensive conversations with insurers throughout the industry, it’s fair to say that insurers are all over the map when it comes to where they are on their AI journey. While we read about many of the AI adoption leaders in the press, most insurers are still in the initial phases of AI adoption. 

Larger insurance organizations may be deep into agentic AI roadmaps, but midsized and smaller organizations are still wading into the generative AI pool in such ways as using Microsoft Copilot or leveraging a GPT model as an AI assistant. Although we see a very large gap between the AI leaders and followers, the adoption pattern is like prior advanced or emerging technologies. However, the pace of change with the advances in AI distinguishes it from past emerging technologies, and organizations that aren’t already getting to “yes” with AI adoption run the distinct risk of being left at a competitive disadvantage.

Surging AI investments

In June 2026, Celent published our 4th Annual AI in Insurance report for the North American insurance market. It shows that insurers in North America rank advanced AI (gen/agentic AI) and large language models (LLMs) as their top technology investment priority. That’s not a surprise, given the energy around gen AI and the unrelenting evolution of agentic AI. 

Four years of Celent market surveys indicate a blistering pace of gen AI adoption — from 8% in 2023 to 28%, 44% and 66% in 2024, ’25 and ’26 — and agentic AI adoption rates are on a trajectory to surpass the gen AI speed of adoption. 

However, all AI matters, and leading insurers are scaling all forms of AI across the enterprise — from DevOps to operations, customer engagement and product capabilities. With so many forms of AI now being used, it’s important that AI capabilities scale consistently. It’s also inevitable that there will be barriers to overcome. 

The chart below highlights the critical concerns insurance organizations have with advanced AI. Not surprisingly, the biggest leap was return on investment (ROI), which went from one of the lowest top areas of concern in 2025 to the top area of concern in 2026. 

The primary reason for this shift in focus is ROI for AI implementations has evolved from a back-burner discussion into a boardroom mandate. A year or two ago, the prevailing AI narrative was driven by a fear of missing out and rapid experimentation. Today, the landscape faces a harsher financial reality.

When capital outlays start reaching today’s scale, the honeymoon phase ends and standard corporate fiscal discipline returns. Finance leaders are no longer satisfied hearing that an employee saved three hours at their function. They want to see that time savings translate into headcount optimization, direct cost reduction or verifiable revenue expansion on the balance sheet.

The sudden intensification of cybersecurity as a primary concern for organizations implementing advanced AI reflects a fundamental change in the nature of the technology itself. Two years ago, security conversations around AI were largely about data privacy and preventing leaks. Today, although data privacy is still a top concern, as organizations are transitioning from text-generating chatbots to autonomous, action-oriented agentic AI, the threat model has entirely transformed.

AI security has officially moved past basic prompt governance and into a strict era of Zero Trust for AI agents, requiring continuous semantic validation and real-time behavioral observability.

Celent insurance survey participants ranked their three leading success factors to realizing the benefits of their advanced AI initiatives. The chart below reflects our 2026 survey results, and for the second straight year, organizational and operational buy-in, data quality/readiness and strong AI governance ranked 1, 2 and 3.

One of the most interesting metrics to highlight is effective training and communication. In 2025, it was only mentioned by 7% of organizations as a top area of focus. Now, approximately 40% of organizations see it as a key success factor, and skilled in-house AI talent has also made an equally dramatic jump.

The pivot toward skilled AI talent, structured training at scale and focused communication on AI strategies has intensified dramatically. A year ago, the corporate landscape was mostly dominated by proof of concepts (POCs) and pilots — organizations were handing out licenses to ChatGPT or GitHub Copilot and waiting to see what happened. Today, that ad-hoc approach has completely broken down. The shift from a passive experimental stance to an aggressive, structured talent strategy is driven by a new reality. 

The chart below indicates that the AI enablers are key areas of focus across all tiers of insurers. These enablers represent the core pillars of what defines a maturing enterprise AI operating model. Attempting to scale advanced AI — especially autonomous, agentic workflows — without these components is how organizations end up with fragmented pilots, runaway cloud costs and severe compliance exposure. 

Compared to 2025, insurers have moved the needle dramatically when it comes to having these AI enablers in place for their AI agenda. In 2025, all of these AI enablers were under 30%. Now, with the exception of having an AI gateway, the AI enablers are at or well over 30%.  

For example, AI governance was only implemented in 28% of insurance organizations participating in our 2025 survey. A year later, the share increased to 81%. AI communication and training plans have been implemented by 63% of insurers – up from 14% in 2025. 

These enabling pillars don’t operate in silos; they form a protective loop around an organization’s technology investments. If one core pillar is pulled out, there’s avoidable exposure. The most important pillar likely is the AI strategy, because it ensures organizations are solving real business problems that move the needle on the bottom line.

Stage 1: Automating everyday tasks

• Advanced AI can be used to automate/augment manual work, typically with a human in the loop.

• Focus on a business problem, realizing low-hanging opportunities using off-the-shelf models and applications.

• Users are also learning, which is a critical first step in developing technical literacy, sparking use cases, building trust and driving adoption.

> Learn more | How Hyland helps insurers innovate and automate 

Stage 2: Changing the nature of work 

• Large-scale productivity gains will require work to be restructured at an organizational or function level with redesigned processes and ways of working. 

• Today, transformations are beginning to occur in functions with high volumes of repetitive, manual, rule-based work (e.g., call centers). 

Stage 3: Reinventing business models and functions 

• Beyond productivity, advanced AI has the potential to reinvent customer experiences, diversify revenue streams and launch new products and business models. 

> Learn more | Maximizing AI’s potential in P&C claims 

The road ahead: Staying at “yes” means measuring KPIs

Measuring success through key performance indicators (KPIs) in operational, financial and customer domains provides organizations with valuable insights and facilitates informed decision-making. Most of our survey participants use operational KPIs as their go-to for measuring success. By leveraging these metrics, organizations can make informed decisions, optimize performance and drive sustainable growth.

Now that most insurance organizations are in production with either gen or agentic AI, KPIs are a key area of focus for tracking realized impact. In Celent’s annual report, operational KPI measurement increased 2x over 2025, and this year’s financial KPI measurement increased more than 3x year-over-year. These increases indicate that as organizations get past their POC and pilot phases, they’re acutely focused on measuring impact for both business benefit and continued investment.

By forcing AI performance to be measured across operational, financial and customer metrics, organizations bridge the gap between data science and business strategy. The benefit of measurement is it converts technical output into boardroom language. 

Instead of saying, “Our LLM has a 92% retrieval accuracy,” you can prove that the LLM reduced submission triage time by three days (operational) and saved $29 per policy (financial), which improved your broker response time and lifted your win rate by 5% (customer). This comprehensive tracking secures long-term executive sponsorship and justifies a continued “yes” to capital investment.

About the author

Keith Raymond is a director in Celent’s North American insurance practice. He has extensive industry experience and is a seasoned expert in AI, process automation, digital transformation and back-office modernization for property and casualty, and life, health and annuities. Keith’s research is focused on process automation through AI across the insurance value chain, and he authors a series of Celent reports on the impact of generative and agentic AI on insurance operations. He has assisted numerous insurance organizations with developing a strategic technology roadmap, go-to-market strategy, core platform replacement selection and IT and operational due diligence on acquisition opportunities. Keith also facilitates Celent’s Gen AI Bootcamp for North American insurers and its annual AI Symposium.