Six roundtables. Six locations. Dozens of voices across carriers, brokers, underwriters technologists and regulators. Over the past eighteen months the commercial insurance industry has undergone a visible and meaningful shift in how it approaches data challenges and the role of Generative AI. What began as fascination with emerging technology has matured into a more grounded and operational understanding of AI’s potential and its limitations.
In April 2024 the first major roundtable we hosted with Insurance Post highlighted a fragmented ecosystem struggling under the weight of complexity. Brokers and carriers spoke openly about the difficulties of driving automation in a world that thrives on exceptions. Manual processes were dominant. Outlook inboxes were the centre of workflow management. Although the value of AI and automation was clear the message was one of caution. The data landscape was messy. Standardisation posed risks of deskilling. And above all the technology needed to serve the client rather than simply improve internal efficiency.
At that stage the idea of augmentation was introduced but largely remained conceptual. Efforts to apply AI were mostly on the margins with transformation leaders trialling tools that supported rather than replaced traditional practices.
In May 2024, sponsored by the UK Innovation Agency, we held a Roundtable in Hartford. All recognised the challenge, accepted it sounded possible to consistently extract data, were keen to experiment and remove the friction of information exchange. Their focus on augmenting the underwriter and specialist, particularly in the Submissions process.
June 2024 – Insurtech Insights New York
June 2024 at ITI in NYC a mix of global carriers and brokers recognised the problem, were excited about GenAI, spoke to their experimentation and its challenges. Finding the edge of what was really possible, understanding the economics, thinking through its application and finding the right use case, commercial line, process and business sponsor was proving difficult and harder than expected.
October 2024 – ITC Las Vegas
October 2024 at ITC in Las Vegas in a panel debate of invited guests the excitement remained, the challenge had moved to mainstream discussion and experimentation. All had initiatives with varying degrees of success. Each starting with a particular use case or document type or process. Few had designed as an enterprise capability so were concerned about expanding beyond the early successes. None had a perspective on long term ROI and their continued investment in what they hoped was a differentiated capability, it was too early for that.
By March 2025, at the ITI Europe roundtable, the tone had shifted. The industry had become increasingly focused on the problem of unstructured data. Participants identified the enormous value locked within documents such as slips, loss runs, bordereaux and ACORD forms. Generative AI showed promise in unlocking this value and alleviating burdensome tasks. However, the enthusiasm was tempered with practical realities. Feeding an entire policy document into a language model was rarely effective. Data orchestration was recognised as essential. So too was the need for validation and governance. The notion of augmentation evolved from a concept to a working principle. AI was not there to replace skilled professionals but to support them at scale.
Different strategic approaches also became more apparent. Large firms were increasingly layering point solutions and building capabilities across their technology stack. Smaller organisations, in contrast, sought more comprehensive solutions that addressed end-to-end challenges without the complexity of bespoke integration. The industry stopped waiting for universal standards. The focus turned to building systems that could adapt to the lack of them.
At the Hartford roundtable in April 2025 urgency replaced theory. Generative AI was delivering value in underwriting support, triage and regulatory compliance. The conversation shifted from what the model could read to what the organisation could do with that insight. The central question became one of control. How could insurers ensure AI outputs aligned with their business logic and internal rules? How could they retain oversight in regulated environments where the firm remained accountable for every decision?
Trust became paramount. Regulators were making it clear that the technology provider or the algorithm itself would not be held responsible. Accountability remained with the firm. As a result, discussions turned to governance, auditability and model explainability. Another key point emerged. Firms must avoid dependency on a single model or provider. In a fast-moving field flexibility is essential. The ability to choose the most suitable model for a particular task and switch between providers protects both innovation and compliance.
The most recent discussion, at the ITI USA roundtable in June 2025, confirmed a new reality. Generative AI is no longer a pilot experiment. It is becoming embedded in operations. The barriers to progress are no longer technical. They are organisational. Integration is the challenge. The leaders will not be those who run the most pilots but those who redesign workflows, augment specialists and retain full control of AI-enabled decisions and have a mature ability to calculate ROI from innovation and make strategic decisions to build for sustainable and structural differentiation and Buy (and customise) for the non differentiating
So what has the industry learned over the past eighteen months?
It has learned that Generative AI is not a silver bullet. It is a powerful tool that demands careful design, robust oversight and ongoing orchestration. The problem of unstructured data remains central. Augmentation is not just the preferred approach but the only sustainable one in a specialist-led market. Decision governance is non-negotiable. Regulatory compliance must be built into AI workflows from day one. And finally, flexibility is a strategic asset. The ability to retain model choice, control intellectual property and avoid vendor lock-in will shape the next era of competitive advantage.
What began as cautious optimism has become confident realism. The commercial insurance industry is no longer chasing artificial intelligence. It is beginning to use it with intent and purpose.
