As talk of “agentic AI” accelerates across the insurance industry, one of the biggest questions emerging is whether autonomy will be a threat or an opportunity for its workforce and distribution channel.
At ITC London 2026, a panel of senior technology and insurance leaders seemed to agree: AI agents are coming to reshape workflows, redistribute value, and ultimately, force insurance firms to rethink their operations. However, human expertise will remain at the heart of these reimagined business models.
According to McKinsey, automation and AI-driven augmentation can already deliver productivity gains of 20-30% in insurance functions such as underwriting and claims, largely by reducing manual data handling.
However, fully autonomous decision-making remains elusive, particularly in complex commercial lines where nuance and judgment dominate outcomes. Sarasola described a “80/20 reality” of AI: today’s AI agents perform well on repetitive, data-intensive tasks, such as extracting information from submissions or pre-populating underwriting files, but still struggle with the final 20% of cases that require contextual judgment or an understanding of ambiguous risk.
The impending shift from automation to autonomous systems will require insurance companies to rethink business models, operating models, and even organisational design. Juan de Castro (pictured on the farthest right), COO of Cytora, a risk digitisation platform acquired by Applied Systems last year, believes insurance is at the cusp of scaling these reimagined business models.
“With any new technology like AI, in the first few years, everyone thinks about how to do exactly what they’re doing today, just slightly better. That’s how AI has mostly been used in the last couple of years,” he said, comparing the current evolution driven by artificial intelligence to technological innovation that upended other industries.
Rolls-Royce, de Castro explained, initially used sensor technology to make engine maintenance cheaper. Then, a few years later, they completely changed the business model: Rolls-Royce owned the engines and rented them based on usage. "This two-phase approach always happens. First, incremental improvement. Then, a new operating model,” he said. “Often, you realise the original processes were broken because they were designed around old technology constraints.”
Concrete use cases for agentic AI are beginning to emerge. De Castro pointed to broker placement workflows, where AI can identify which carriers are most likely to quote a risk and automate much of the back-and-forth exchange of missing data and clarifications. In theory, this allows brokers to approach more carriers without additional effort, potentially improving outcomes for clients.
At WTW, Sarasola highlighted how large language models are accelerating long-standing ambitions around machine-readable policies and computable contracts. Tasks that once took months can now be performed dynamically, she said, opening the door to deeper integration across pricing, claims, and reserving.
However, the panel was cautious about overstating AI’s near-term capabilities.
“The levels of abstraction and different types of (sensory) information are not part of these (large language) models yet,” Sarasola said. “There are a lot of technical hurdles to doing that effectively, and it might not happen in the next 10 years.”
Cultural and talent challenges may be even harder than the technology itself. With large portions of the insurance workforce approaching retirement, panellists stressed the importance of capturing institutional knowledge while reassuring professionals that AI is a teammate, not a replacement.
Recent insurance surveys suggest the biggest barriers to AI adoption are often people-related (i.e., skill/resource constraints and resistance to change) alongside persistent data challenges.
“We can have phenomenal technology in this industry, but culture and talent are what determine whether that technology actually gets used, and used in the right way,” said Scott Sayce, chief innovation officer at DUAL Group (pictured middle, left). “For us to succeed as a market, we have to be far more intentional about how we take talent on the journey.”
Sayce stressed that insurance firms must carefully balance talent and technology: “Yes, we need to move fast, but if we move too fast, we risk losing talent.”
