Sona’s chief operating officer did not set out to build a career in insurance, yet nearly two decades on he is helping steer a fast-growing brokerage through AI adoption, niche digital products and a changing claims landscape, while insisting that people, not technology, remain the point.
James Potter (pictured) did not arrive in insurance through a carefully laid plan. He finished education and was pointed towards the sector by a family member already working in it. What kept him there was its range: an industry, he says, made up of “an interesting mix of legal, sales and everything”.
That early suggestion has since turned into a career of about 18 years, split broadly between broking and underwriting, giving him what he describes as experience on both sides of the market. He began at Lockton on specialist high-net-worth motor schemes, including Ferrari and Aston Martin owners’ programmes, before moving to Marsh, where he worked with wealthy private clients, farms and estates, vehicle collections and Lloyd’s placements. He later crossed into underwriting at Hiscox, taking on what he said had been one of the market’s biggest risks, before periods at Towergate, NFU Mutual and Gallagher which led him to Sona.
That progression matters because it helps explain the perspective he now brings to operations. At Sona, he says, he has come to understand insurance “from farm to table”, spanning front-end distribution, underwriting logic, finance, cash and compliance. He joined when the business had only a handful of clients; more than three years later, he says, it has grown to a couple of thousand clients, with millions in gross written premium in two locations.
Potter’s account of digital transformation is notably unsentimental. For Sona, he says, it is simply “future proofing the business”. In practice, that has meant pursuing two tracks at once. AI enablement inside the brokerage, and digital products for customers outside it. The firm remains, in his description, a conventional manual broker built around phone calls, emails and personal client contact. But it is also seeking areas where specialist products can be bought quickly online, without routine human intervention.
The internal case for AI was not built around novelty. Potter says the point was to remove “manual, lengthy processes” that absorb time without adding much value, thereby releasing staff for the higher-value work of dealing with clients. One example was the production of repetitive “to whom it may concern” letters and similar cloned documents, the sort of task that can consume an hour and a half without materially improving the client relationship.
Sona’s route into AI came through the Willis network of independent brokers, where conference discussions and external speakers appear to have given the company its initial push. From there, the business reviewed large language models and opted for ChatGPT with an enterprise subscription for all staff.
For a regulated intermediary, however, AI implementation quickly becomes a question of control. Potter says compliance was Sona’s “first hurdle”, and that the business adapted a risk-management framework with help from Willis. The test was pragmatic: compare the consistency of machine output against the notional human error rate, and keep a human final check in place.
That caution extends especially to data. Sona is still using AI in a relatively manual way and, Potter says, does not trust the environment enough to input personal data. Staff are instructed to use the company’s approved enterprise subscription rather than their own tools, and “under no circumstances” to submit personal information. The rationale is straightforward enough: enterprise plans may promise not to use customer data for model training, but the wider behaviour of such systems remains insufficiently understood for a broker handling sensitive client material to be relaxed about it.
What is striking is that Potter does not present AI as a cost-cutting exercise in the crude sense. Sona’s phrase is not “AI first” but “AI forward”, and he is explicit that the aim is not staff replacement. Instead, he describes a capacity model: once a team reaches full stretch, management’s job is to use people or systems to bring it back down to a sustainable level.
One of the more interesting points in Potter’s interview is that Sona sees AI as an inclusion tool as well as an efficiency tool. He argues that for staff with dyslexia or ADHD, AI can provide immediate support with rereading, checking, grammar and common errors, reducing the need to rely on colleagues and helping them operate “on a level playing field”.
That matters in an industry that, as Potter notes, faces a demographic problem. He points to repeated warnings about retirement-led attrition over the coming decade and argues that AI may help widen the available talent pool by making more roles accessible to people with strong interpersonal or analytical strengths who may otherwise be disadvantaged by conventional written processes.
It is also, he suggests, easier to secure adoption when the pitch is made honestly. Sona is a relatively young company, with staff already familiar with AI in daily life, though not always through approved tools. Potter says younger employees were often quick adopters, while others needed more persuasion. The message to both groups was that this was “an investment in you guys”, not a precursor to redundancy, and that refusing to engage with such tools would leave them behind in the next phase of work.
Like much of the broking market, Sona still depends on Acturis. Potter calls it the dominant platform in the sector, with the best e-trade facilities and the stability that comes from being the established market choice. Yet he also portrays it as a brake on change, arguing that broking will move materially faster only when AI functionality is embedded in the core system itself. For now, Sona’s Acturis set-up is cloud-based, but not especially intelligent.
That creates a familiar tension. Brokers may want to accelerate, but their infrastructure often ties them to systems built for robustness rather than invention. Potter’s answer is not to wait. Sona has tried to push innovation into adjacent areas, particularly through niche digital products.
The most developed of those products is ComicSure, a pre-launch collectors’ insurance proposition developed under the Sona umbrella but with its own branding. The origin story is unusually personal: Sona’s chief executive, Rob Thacker, is described by Potter as a serious comic-book collector. Yet beneath the colour sits a sensible underwriting point. Comic books are specialist assets whose value is not easily captured by standard household insurance processes, and Sona believes there is room for a product that relies on knowledgeable valuations rather than generic receipts.
Potter says Sona has spent about 12 months working on the proposition and has drawn heavily on the University of Essex business community around it, using nearby marketing firms, data specialists, AI companies and Insurtech contacts to help shape brand, systems and capacity. That ecosystem appears central to the firm’s approach. He speaks more than once of being able to “walk down the corridor” to find expertise, and credits the surrounding network with helping Sona’s leadership get up to speed on change.
A second initiative is earlier in development but arguably speaks to a more immediate customer problem: an unoccupied-property probate product. Potter says the target customer is the executor or solicitor trying to manage a property after a death, at a time when insurance is usually one more distressing administrative burden. Sona began work on the product roughly six weeks before our interview, after identifying a gap left when a larger broker withdrew from an existing arrangement. Potter says the business now hopes to launch something this year.
This is where Potter’s broader philosophy becomes clearer. Sona is not attempting to become a mass-market digital insurer. It is looking for specialist spaces where expertise, service and narrow product design can combine into something commercially useful.
If Potter sounds sceptical of grand transformation rhetoric, he is equally sceptical of leaders who sponsor change they do not understand. His view is that digital transformation “can only really take place when the leadership understands it”, and that senior teams have to go first: learn it, use it, and explain it. Otherwise, change programmes stall in the gap between aspiration and comprehension.
That argument will resonate in the commercial insurance market, where many firms remain caught between top-down innovation language and bottom-up practical constraints. Potter’s point is less fashionable than some AI evangelism, but perhaps more useful: transformation succeeds only when it is translated into how people actually work.
Looking ahead, Potter is especially interested in the rise of AI agents in claims. His concern is not that they are impossible, but that the claims journey often depends on interpretation, context and specialist advice rather than simple task execution. He wonders whether insurers will confine AI to first notification of loss or push it deeper into the process, and whether that, in turn, may strengthen the broker value proposition for clients who do not want to argue a complex interruption or coverage issue with a machine.
That is a telling conclusion. For all the emphasis on automation, Potter’s instincts remain firmly brokerish. Technology may alter workflows, improve consistency and generate niche products. But his underlying thesis is that when risk becomes messy, clients still want somebody in their corner.
Outside work, Potter presents himself less as a digital evangelist than as a family man. He is father to two daughters, Thea and Sienna, and says Sona’s family-friendly culture has allowed him to be deeply involved in their lives.
He is also midway through an MBA at the University of Wolverhampton, a return to formal study after deciding as a younger man that he preferred work to university. Spare time, he suggests, is limited. Much of what remains is spent with his family, often at Colchester Zoo. In an interview mostly about AI, risk and distribution, it is an ending that feels properly human.
