Some thoughts on the potential of AI, making savings and what insurers should spend those savings on, from Gleb Tsipursky, PhD, a behavioral scientist, CEO of Disaster Avoidance Experts – full bio below;
Insurance is an apprenticeship business disguised as a data business. Junior claims handlers and underwriters learn by seeing ordinary cases, then discovering the details that make some of them extraordinary. AI can speed the ordinary work. It should also speed the learning.
Stanford’s August 12 employment update analysed U.S. payroll data through June 2026. Employment among workers ages 22–25 in highly AI-exposed occupations was about 19% below the path it would have followed if it had kept pace with similarly aged workers in less-exposed occupations. The comparable gap was 15% in the July 2025 data vintage. The adjustment appears mainly through reduced hiring, especially in occupations where AI tends to automate human tasks.
Insurers have plenty of work that looks ideal for automation: first-pass claim summaries, policy comparisons, document extraction, fraud flags, customer correspondence and routine underwriting preparation. The productivity case is obvious. The capability risk is less visible.
A new claims professional who never builds a first chronology can miss how facts change meaning as a file develops. A junior underwriter who receives an AI-generated risk summary can become fast at accepting conclusions before learning how to test them. A customer-service employee who relies on generated explanations can struggle when a policyholder’s situation falls outside the standard path.
Insurance Edge describes its audience as focused on digital transformation and innovation across claims, fraud and insurtech, and its 2026 editorial programme invites direct industry comment. That makes the next stage of AI adoption an operating-design question, not a tool-selection question.
The answer is to automate preparation and deliberately increase exposure to judgment. If AI summarises a claim, give the junior handler responsibility for identifying contradictions and missing evidence. If it proposes an underwriting rationale, require the employee to test the assumptions against policy language and loss history. If it drafts a customer explanation, have the employee handle the cases where emotion, ambiguity or vulnerability changes the conversation.
Skills England’s 2026 guidance on AI upskilling stresses practical, role-specific learning tied to real work. Insurance leaders can apply that directly by pairing automation with structured exception practice and experienced review.
Then measure the result. Alongside cycle time and cost per claim, track time to independent competence. How long before a new handler can own a non-standard case? How long before a junior underwriter can defend a risk decision? How quickly can an employee recognise when an AI-supported answer needs escalation?
This changes the economics of automation. A system that saves minutes but leaves experienced staff doing all the difficult thinking creates a bottleneck. A system that saves preparation time and moves junior employees into supervised exceptions builds capacity.
Insurers need experienced judgment years from now. The fastest way to get it is to design AI around the career ladder rather than around the deletion of routine tasks. Automate the paperwork. Keep the learning. Use the saved time to make people competent sooner.
Gleb Tsipursky, PhD, a behavioral scientist, CEO of Disaster Avoidance Experts, and author of The Psychology of AI Adoption at Work: From Resistance to Results (Georgetown University Press, 2026). https://disasteravoidanceexperts.com/aibook
Be the first to comment