For an industry inundated with AI pilots, insurance remains conspicuously slow to bring artificial intelligence into full-scale production. According to Vikas Acharya (pictured), CEO and co-founder at ChainThat, a key reason AI projects stall lies in a mismatch between technological potential and organisational readiness.
"Proof-of-concept is often about proving the technology, not the value," said Acharya. "People get excited about what AI can do but forget to define what success looks like. Without clear KPIs and executive sponsorship, these projects tend to lose steam."
While early blockchain initiatives offer a historical parallel - many failed to transition beyond experimentation - AI faces its own unique barriers. Chief among them is culture.
"There's nervousness," Acharya said. "Some insurers ask, 'Should we lead, or should we follow?' That hesitation becomes a blocker."
AI pilots often focus on the wrong problems, he noted, or expend disproportionate effort on tasks that don't reflect the actual business challenge. "You're solving 20% of the problem with 80% of the effort. And often it's not even the right problem."
Security and data privacy add additional friction. "As insurers begin to sell AI-based products, liability concerns come into play. I've seen cases where cyber cover doesn't extend to AI-driven exposures. That creates a whole new risk landscape."
Resistance also arises at the operational level, particularly in underwriting. Acharya points out that AI is often perceived as a threat to professional judgment.
"Underwriters pride themselves on domain expertise," he said. "Turning that into a black box algorithm can feel like you're stripping away their role."
To be accepted, AI must be framed as an augmentation tool, a collaborator rather than a replacement. "It has to respect niche knowledge and enable better decisions, not override them."
On the systems side, data integration remains the most persistent legacy challenge. "Decades of modernization have left insurers managing layers of legacy tech," said Acharya. "Most of the effort goes into syncing systems, which inflates the cost and risk of AI deployment."
One promising AI application involves using language-based interfaces to streamline policy setup. In a recent implementation, an insurer used AI to automate the configuration of insurance products based on natural language inputs from business analysts. This reduced manual setup time and allowed faster product deployment without interfering in underwriting decisions.
This approach worked, Acharya noted, because it targeted a clearly defined process and respected the boundaries of professional judgment.
For brokers, AI adoption within carriers isn’t just a back-office story, it can directly impact service quality. Faster policy setup and improved workflow automation can translate to quicker quotes, fewer delays, and a more responsive customer experience.
Understanding how and where insurers are deploying AI helps brokers set expectations with clients, especially in complex or regulated lines. It also offers insight into which carriers may be better positioned to handle bespoke risks or rapidly changing requirements.
When asked whether regulation is a major blocker, Acharya demurred. "It depends on the use case," he said. "If it's about guiding and assisting users, regulation isn't a big hurdle. But if it's involved in decisions that affect coverage, then yes, the regulatory bar is higher."
Acharya advises firms to avoid over-engineering. "A perfect solution might not be the right one for your constraints," he said. "Budget, geography, data sensitivity, you have to work within your context to build trust without overcomplicating things."
Looking ahead, Acharya sees promise in tools like GPT-5, particularly in their ability to simulate scenarios and interpret context. "It helps underwriters think differently," he said. "They can ask 'what if' questions and test assumptions, which supports better risk selection and pricing."
Still, he cautions that advanced AI is not a silver bullet. "It all comes down to use case," he said. "If you understand your problem and apply the right tool, AI can create real value. But if not, it’s just another stalled project."
For insurers hoping to go beyond the pilot phase, that clarity may be the most important innovation of all.
