Turning AI initiatives into commercial outcomes: lessons from reinsurance

At VISIONS’ Data Strategy and Analytics Summit on 21 April 2026, Ahmed Al Mubarak, Director of AI and Data Science at Howden Re, spoke on the AI Monetisation and New Business Models panel. During the panel, Ahmed and his fellow participants sought to address a central question for organisations investing in AI: where are they seeing real returns? 

Ahmed’s comments focused on a consistent gap between technical progress and commercial outcomes. While AI capability continues to advance, many initiatives struggle to translate into measurable business value. He argued that the most common issues occur in how AI is applied within industry workflows, how trust is built with users, and how solutions are delivered at scale. In reinsurance, this distinction is already shaping which organisations are moving beyond experimentation and into sustained value creation.

Why general-purpose AI falls short in reinsurance

A focus of the panel was the limitations of general-purpose AI tools. Most organisations already have access to foundation models and widely-available agents. As a result, the incremental value of another generic capability is limited. Ahmed argued these models are not designed to understand how a reinsurance contract is negotiated, how broker submissions are structured, or how placement decisions are made. Without that context, outputs can lack specificity and introduce risk in decision-making. 

Ahmed stated, “The opportunity lies in building AI that is embedded in industry workflows and aligned to real user needs. The model is the easy part. Distribution is where the money is made or lost.”

From capability to execution

Moving from capability to commercialisation requires more than technology. Ahmed outlined the three foundations to the audience: technology, data, and operating model. While architecture continues to evolve towards more modular and domain specific systems, and data readiness remains a challenge due to unstructured sources, he argues the operating model is often overlooked. 

“Clear accountability, effective change management, and the ability to connect AI outputs to revenue are critical,” he noted. “Without these elements, you do not have a commercial AI capability, you have a research project.”

Trust, cost and adoption

Ahmed went on to discuss how cost and trust are closely linked in shaping adoption. While attention often focuses on the expense of running large models, he highlighted that the greater cost comes from uncertainty around when to rely on AI outputs. 

This is driving a shift towards smaller, more efficient models for targeted tasks. At the same time, human oversight remains important, not primarily for regulatory reasons, but to build confidence among users and clarify accountability. “Human-in-the-loop is not a governance decision, it is a trust decision,” he explained.

Measuring value and building advantage

Panellists argued that measuring value presents a further challenge. Technical metrics such as accuracy are necessary but insufficient. Behavioural indicators, including whether users adopt the tool and change their decisions, provide a clearer signal of impact. Ultimately, commercial outcomes are the most difficult to attribute, particularly in complex reinsurance transactions, emphasised Ahmed. “Without a clear link between AI outputs and business decisions, value remains unproven. A 95% accurate model that nobody uses creates zero value.”

Looking ahead, panellists agreed that the ability to scale AI will depend on how effectively organisations build distribution into their strategy. Embedding AL into existing platforms, partner ecosystems, and client workflows will define long term advantage. As Ahmed concluded, “In five years, the winners will not be the organisations with the best models. They will be the ones with the best distribution.”