Agentic AI is moving into the workplace. Here’s what actuaries need to know about working alongside a new kind of digital colleague.
By Beth Baker
Throughout history, technological shifts have transformed the way we work. Desktop computers replaced manual table lookups, spreadsheets automated complex calculations, and cloud computing enabled unprecedented access to data. With every major technological shift, innovations initially sparked skepticism before widespread adoption, ultimately enhancing rather than replacing the actuary’s essential role. Today, we stand at another inflection point: the emergence of agentic artificial intelligence (AI).
While actuaries have long used statistical models and software tools and are gaining familiarity with generative AI, agentic AI represents something different. These systems are capable of autonomously pursuing goals, making decisions, and iteratively solving problems with minimal human intervention. The insurance industry continues to embrace this technology. According to PwC’s “Insurance Executives Say Agentic AI Is Leading Industry Transformation,” 57% of insurance executives prioritized both generative and agentic AI as top technology investments in 2026. Actuaries must prepare for significant changes in how we work.
What Is Agentic AI?
Agentic AI refers to artificial intelligence systems that autonomously pursue goals and make decisions without constant human direction. Their capabilities outpace traditional AI tools, which require explicit programming or prompting for each task, and surpass traditional automation, which relies on rigid rule-based algorithms. Agentic AI can plan, execute, and adapt its approach based on feedback and changing circumstances. Additionally, these systems respond to natural-language inputs, eliminating the need for users to code to make them function.
The key characteristics that define agentic AI include autonomy in decision-making, goal-oriented behavior that persists across multiple steps, and iterative problem-solving that learns from outcomes. Think of it as the difference between a calculator that performs operations when you press buttons and an assistant who understands your goal to analyze reserve adequacy. The calculator is a useful tool, but the assistant independently gathers data, runs scenarios, identifies anomalies, and presents findings, allowing you to focus on other priorities until its work is complete and ready for review.
These abilities represent a fundamental shift from assistive to autonomous AI. AI agents exist on a spectrum from augmentation (assistants that enhance productivity) to automation (handling single- or multistep workflows) to fully AI-native enterprises that weave intelligent systems throughout their operations. For actuaries, this means technology that doesn’t just speed up existing processes; agentic AI can fundamentally reimagine workflows.
Agentic AI in Practice
To understand agentic AI’s potential, consider applications across industries. Customer service agents now resolve complex issues end-to-end, both diagnosing problems and coordinating solutions across multiple systems. Research assistants autonomously gather information from diverse sources, synthesize findings, and produce analytical reports. Software development agents like Devin.ai write, test, and debug code with minimal human oversight.
In financial services, AI agents investigate market trends, identify anomalies, and generate detailed analysis reports. What makes these systems “agentic” isn’t their sophistication alone but their ability to operate autonomously toward defined goals, making intermediate decisions and adapting strategies as needed.
From an actuarial perspective, agentic AI promises significant benefits across multiple dimensions of practice. Industry experts, including projections from McKinsey and others, indicate that insurers deploying AI agents in core functions such as claims processing could reduce processing time by as much as 70% to 80% for routine claims. Actuarial data preparation tasks that once took days can now be completed in hours through automated data extraction and validation.
Time Savings and Enhanced Focus
The most immediate benefit of agentic AI is operational efficiency and time reclamation. Gartner predicts that by 2029, 80% of common customer service issues will be handled autonomously by agentic AI. For actuaries, AI agents that handle routine tasks such as data extraction, report drafting, data visualization, and documentation free up an actuary’s time to perform more complex tasks. Instead of spending hours compiling mortality rates, claims frequencies, and policy counts, actuaries receive clean, structured data ready for modeling. This shift allows us to focus on what truly requires actuarial judgment: evaluating assumptions, considering emerging risks, and providing strategic guidance.
Consider statutory filings, which traditionally require pulling data from multiple sources and formatting according to annually changing requirements. An AI agent can execute this process, apply current reporting rules, and flag items for actuarial review. What once took a team three weeks might now take three days, with actuaries spending time on exceptions rather than routine calculations.
Enhanced Accuracy and Pattern Recognition
Agentic AI systems excel at consistency and can identify patterns in massive datasets that might escape human notice. In claims analysis, AI agents can process thousands of first notices of loss, extract key data points, categorize severity, and flag unusual patterns for investigation. This enhanced pattern recognition can improve fraud detection, risk identification, risk segmentation, and underwriting functions, enabling more precise pricing and reduced adverse selection.
Professional judgment about assumption appropriateness, consideration of emerging risks, ethical implications of pricing decisions, and communication with stakeholders all require distinctly human capabilities.
For pricing reviews requiring detailed experience studies across multiple dimensions, such as coverage type, geography, industry, and policy size, AI agents can break down data and perform preliminary analysis, allowing actuaries to focus on the interpretive work: Are the patterns we’re seeing meaningful? Do our trend factors remain appropriate given market changes? Should we adjust our segmentation approach?
Understanding reinsurance programs requires reading treaty wordings, tracking cessions, and modeling various loss scenarios. AI agents such as Nomad Data’s Doc Chat can extract treaty terms from agreements, identify coverage gaps or overlaps, and simulate how different loss patterns would impact net retention. This capability proves particularly valuable for complex programs with multiple layers and facultative placements, freeing actuaries to focus on strategic decisions about risk transfer and capital optimization.
The Critical Partnership: Human-in-the-Loop
Despite agentic AI’s impressive capabilities, fields requiring professional judgment benefit from keeping humans in the loop to actively oversee, guide, and validate the system’s decisions. Human oversight may be required to maintain regulatory compliance and public trust. As a comparison point, think of AI agents functioning like entry-level staff who facilitate processes while experienced actuaries maintain supervisory authority.
Human oversight operates at different levels depending on the task’s risk and complexity. For routine data extraction and formatting, monitoring with periodic spot-checks may suffice. For pricing decisions or reserve analyses, human approval before implementation becomes critical. For complex strategic decisions involving multiple assumptions and significant financial impact, deep human involvement throughout the process remains essential.
The actuarial profession’s fundamental responsibility—
protecting the public interest through sound risk assessment—cannot be delegated to algorithms. Professional judgment about assumption appropriateness, consideration of emerging risks, ethical implications of pricing decisions, and communication with stakeholders all require distinctly human capabilities. AI agents should augment these capabilities, not replace them.
Building effective human-AI workflows requires clear delineation of responsibilities, robust feedback mechanisms, and continuous evaluation. Like managing human staff, this includes setting clear expectations, providing guidance on edge cases, reviewing outputs systematically, and adjusting the system based on performance. The most successful implementations treat AI agents as team members requiring onboarding, job descriptions, and ongoing development.
Essential Safeguards and Risk Management
The deployment of agentic AI in actuarial work demands rigorous safeguards. Recent Gartner research suggests that up to 40% of early agentic AI projects may be abandoned by 2027, often due to inadequate governance and risk management frameworks. Successful implementations require multiple layers of protection and direction, including governance, compliance, risk controls, technical safeguards, monitoring, auditing, and strategic implementation.
Model validation and output verification protocols must be established before deployment. This includes testing AI agents against known scenarios, comparing outputs to traditional methods, and establishing clear thresholds for acceptable performance. Due to their complexity, AI systems can become black boxes, preventing the transparency that’s valuable in actuarial work. This tension demands enhanced validation procedures, not abandonment of technology.
Bias detection and fairness testing require particular attention. AI systems can inadvertently perpetuate or amplify biases present in training data, potentially leading to discriminatory outcomes in pricing or claims handling. Regular audits examining outcomes across demographic groups, geographic regions, and policy characteristics are essential. Industry observers were expecting regulatory scrutiny of AI fairness to intensify significantly this year.
Comprehensive audit trails and explainability requirements ensure accountability. Every AI agent decision should be traceable: what data was used, what logic was applied, what alternatives were considered. This documentation serves multiple purposes: regulatory compliance, error diagnosis, continuous improvement, and stakeholder communication. The European Union’s AI Act and similar regulatory frameworks increasingly mandate such transparency.
Data security and privacy protections cannot be overlooked. AI agents often require access to sensitive information, and their autonomous nature creates new attack surfaces. Robust access controls, encryption, and monitoring for unusual activity are essential. The insurance industry has seen cyber threats specifically targeting AI systems, making this area of particular concern for 2026 and beyond.
Finally, clear error handling and fallback mechanisms must be designed upfront. What happens when an AI agent encounters unexpected situations? How are errors escalated? What manual processes can continue if AI systems fail? These considerations, while unglamorous, prove critical to maintaining operational resilience.
Why Actuaries Need to Be Aware Now
The pace of AI advancement demands immediate attention from actuaries. ChatGPT reached 100 million users within two months, outpacing the adoption curves of both the internet and personal computers. Within two years of ChatGPT’s launch, generative AI achieved a 39.5% adoption rate, a milestone that took the internet a decade to reach. A Celent survey found that 22% of insurers expect to have agentic AI in production by the end of 2026, with many industry observers believing the actual number will be higher.
Early adopters will gain significant competitive advantages. Organizations implementing AI agents across actuarial functions may experience substantial improvements in analytical capacity, faster response to market changes, and enhanced ability to identify emerging risks. These advantages compound over time as organizations refine their systems and build institutional knowledge around AI deployment.
Professional responsibilities and standards are evolving alongside this technology. Actuaries must understand AI capabilities and limitations to fulfill their professional obligations effectively. This includes being able to evaluate AI-generated analyses, understand when human judgment is essential, and communicate both the strengths and limitations of AI-augmented work to stakeholders.
Client and stakeholder expectations are shifting rapidly. As AI adoption accelerates across industries, clients will expect faster turnaround, more sophisticated analyses, and deeper insights. Organizations that leverage AI agents to meet these expectations while maintaining actuarial rigor will have significant advantages in client relationships and market positioning.
The regulatory landscape is changing as well. Regulators worldwide are developing frameworks for AI governance, with implications for transparency, fairness, and accountability. Actuaries familiar with these frameworks and experienced in implementing compliant AI systems will be valuable resources for their organizations.
Challenges and Considerations
Despite the compelling benefits, significant challenges accompany agentic AI adoption. The learning curve for actuaries and their organizations should not be underestimated. Understanding AI capabilities, limitations, and appropriate applications requires investment in education and training. Our professional organizations are responding actively: the Casualty Actuarial Society (CAS) offers programs such as the AI Fast Track On Demand; the Society of Actuaries (SOA) provides AI-focused courses, certificates, symposia, and bulletins; the American Academy of Actuaries (Academy) publishes professionalism guidance, develops public policy
resources and comment letters, and hosts educational webinars through its AI, Data Science, and Analytics Committee; and the International Actuarial Association (IAA) advances global competency through its AI Task Force, educational resources, governance frameworks, and the recently launched website, aiforactuaries.org.
Integration with existing workflows and legacy systems presents practical difficulties. Many actuarial teams still rely on spreadsheets, proprietary databases, and aging policy administration systems. Connecting AI agents to these systems, ensuring data quality, and maintaining operational continuity during transition require careful planning and execution.
Cost considerations extend beyond software licensing. Successful AI implementation demands investment in computing and data infrastructure, technical expertise, change management, and ongoing monitoring and refinement. Organizations must evaluate whether these investments align with their strategic priorities and whether they have the resources to sustain long-term AI operations.
Governance and privacy regulations add complexity. Different jurisdictions impose varying requirements on AI use, particularly regarding personal data and algorithmic decision-making. Multinational insurers must navigate this patchwork of regulations while maintaining consistent operational standards. In the United States, regulations vary by state, creating additional complexity.
Managing stakeholder trust and transparency becomes more complex with AI systems. Explaining how AI agents contribute to actuarial conclusions, maintaining confidence in results, and addressing concerns about algorithmic bias or errors require thoughtful communication strategies. The actuarial profession’s credibility depends on maintaining transparency about methods and limitations.
Perhaps most importantly, actuaries must maintain professional skepticism. AI agents can produce confident-sounding results that are nonetheless incorrect, the “hallucination” problem in generative AI. Critical evaluation of AI outputs, cross-checking against alternative methods, and questioning unexpected results remain essential actuarial skills.
The Future of the Actuarial Profession
The integration of agentic AI signals evolution, not obsolescence, for the actuarial profession. History provides reassurance: every technological revolution results in displacement, transformation, and new roles. Desktop computers didn’t replace actuaries; they freed them from mechanical calculations to focus on more sophisticated analyses. Cloud computing didn’t eliminate the profession; it enabled actuaries to work with larger datasets and more complex models.
Agentic AI will similarly expand the actuarial role rather than diminish it. As AI handles routine data processing and preliminary analyses, actuaries can dedicate more time to strategic thinking, complex risk assessment, ethical considerations, and stakeholder communication. The profession will shift toward higher-value activities requiring judgment, creativity, and deep domain expertise.
This evolution will require new skills. Actuaries of the future will need to understand AI capabilities and limitations, evaluate AI-generated analyses, design effective human-AI workflows, communicate about AI systems to technical and nontechnical audiences, and navigate evolving regulatory frameworks around AI. Technical fluency with AI systems, while not requiring deep programming expertise, will become as fundamental as spreadsheet proficiency is today.
Yet the enduring value of actuarial expertise remains unchanged. AI cannot replicate professional judgment informed by years of experience, ethical reasoning about fairness and equity in insurance systems, ability to identify novel risks not captured in historical data, communication skills for explaining complex analyses to diverse audiences, and understanding of the broader business and regulatory context. These distinctly human capabilities will remain central to actuarial value.
Preparing for this AI-augmented future requires proactive engagement. Actuaries should seek opportunities to work with AI systems in low-risk environments, participate in professional education programs, contribute to developing best practices and ethical guidelines, advocate for responsible AI adoption within their organizations, and maintain the critical thinking skills that distinguish human judgment from algorithmic processing. For those in leadership roles, providing a clear adoption strategy, demonstrating the benefits of use cases, and offering a supportive skill development program will alleviate staff concerns about job loss.
The actuarial profession stands at a unique vantage point. Our training in probability, statistics, and risk management positions us well to understand the capabilities and limitations of AI systems.
An Opportunity to Lead
Agentic AI represents a transformative force in actuarial practice, offering unprecedented capabilities for data analysis, pattern recognition, and workflow automation. This technology promises substantial benefits: time savings through automation of routine tasks, enhanced accuracy and consistency, ability to analyze larger and more complex datasets, improved risk identification, and more time for strategic work and client interaction.
However, realizing these benefits requires careful attention to essential safeguards: robust validation protocols, bias detection and fairness testing, comprehensive audit trails, data security and privacy protections, and clear error handling mechanisms. The human-in-the-loop principle must remain central, ensuring that professional judgment guides AI deployment and that actuaries maintain ultimate accountability for their work.
Our role in this transition is essential: Embrace the possibilities, prepare for the future, and adapt our role accordingly. Begin exploring AI capabilities through pilot projects and professional education. Develop frameworks for evaluating AI systems and integrating them responsibly into actuarial workflows. Contribute to the profession’s collective understanding of how to leverage AI while maintaining actuarial standards and ethics.
The actuarial profession stands at a unique vantage point. Our training in probability, statistics, and risk management positions us well to understand the capabilities and limitations of AI systems. Our professional standards emphasize transparency, fairness, and public protection, aligning naturally with principles of responsible AI adoption. Our experience navigating technological changes throughout history demonstrates our profession’s adaptability.
Your next coworker may indeed not be human, but that colleague will work under your supervision, follow your guidance, and augment rather than replace your professional judgment. By engaging thoughtfully with agentic AI now, actuaries can shape its development, establish best practices for its use, and ensure it serves the profession’s fundamental mission: protecting the public interest through sound risk assessment.
The opportunity to lead in the responsible adoption of AI is ours to seize.
Beth Baker, MAAA, FSA, FCA, MBA, is founder and chief actuary of Naxara.
