After reading mathematics and computer science at Imperial College London, Vincent Liu’s (pictured) first position out of university was at Acturis, the UK’s largest broker software house, where he spent two years leading internal improvement projects. The experience gave him an unusually clear view of the sector’s structural pressures, the weight of legacy systems, and the friction that slows the broking process at every level.
What followed was a deliberate programme of self-education through scale. Liu moved into property technology and a series of early-stage startups, joining each as the first engineering hire and building products and platforms from scratch. “Those experiences gave me a pretty comprehensive view of how running a business looks,” he explained, “not just at a large enterprise but also at a small startup, and how it evolves into a scale-up and then into an enterprise.”
By the time Liu felt ready to build something of his own, he had developed ideas around making life easier for software engineers. He built AI products, achieved promising traction, and then concluded that the unit economics could not sustain the model. The pivot that followed would prove more consequential. On an online platform, he encountered Mark and Jake, who would become his co-founders at Meshed. Their backgrounds were complementary in precisely the ways that mattered: Liu bringing technical AI capability and insurance knowledge, Jake contributing ten years of insurance product experience, and Mark providing hands-on broking expertise.
Their founding vision, in Liu’s telling, was straightforward: “build an insurance broking experience that businesses love and make it super-efficient so we’re not overcharging in commission,” he said.
Meshed was formally incorporated May 2024. Full FCA authorisation followed in November 2025.
The commission question is central to Meshed’s proposition and perhaps its most immediate commercial differentiator. Across the commercial insurance market, commission rates vary considerably by product line. Fleet, for example, tends to sit between ten and thirteen per cent. For the majority of commercial lines, however, the industry can charge up to forty per cent.
Meshed charges ten per cent across its book.
Liu is direct about why that matters to clients: “They only see insurance as a cost. They don’t necessarily see it as something they are paying to protect their business. It’s always a cost to them,” he said. By reducing commission to a fraction of what many brokers can charge, Meshed argues that it can deliver material savings while remaining profitable, an efficiency made possible by the extent to which AI handles what would otherwise be labour-intensive processes.
The company targets businesses with annual revenues above £1 million and up to £30 million, a segment that sits awkwardly between the fully standardised consumer market and the bespoke arrangements available to large corporates. Comparison websites and instant-quote platforms have served the small end of the market well; Meshed is betting that a different kind of technological solution is required for the mid-market.
The choice of commercial insurance as Meshed’s operating territory was not arbitrary. Liu identifies the structural complexity of the segment as precisely the feature that makes it amenable to AI intervention. Below the £1 million annual revenue threshold, commercial insurance broadly resembles its consumer counterpart: standardised question sets, templated underwriting, and commoditised pricing. Above that threshold, the picture changes fundamentally.
“AI can standardise in a way that makes broking almost uniform,” Liu said, “because AI is dealing with the unstructured uncertainty of the process.” Underwriters serving mid-market businesses tailor their questions to specific operating models; there is no single template, no unified question set shared across the market. It is precisely that absence of standardisation that Liu believes AI can address, extracting structure from an otherwise fragmented process.
The competitive landscape reinforced the decision. Existing digital brokers have concentrated their efforts on sole traders and micro-businesses, where the economics of instant-quote technology work well. Meshed’s research identified a gap above that level. “Companies with more than £1 million ARR are underserved in the industry,” Liu observed. “No brokers have created a solution for them yet. It’s all phone calls and emails.” That is where the company has chosen to operate, and where it reports the strongest product resonance.
Meshed’s decision to build its entire technology platform in-house rather than adopt established systems such as Acturis is both a practical and philosophical position. Liu describes the gravitational pull of legacy systems with a directness that reflects his own experience: “Once you’re stuck with them, it is very difficult to get out. It’s almost like gravity,” he said.
The problem, as he sees it, is not merely that legacy systems are inflexible. It is that the companies behind them have little incentive to expose data or connections to external platforms, which effectively locks clients in and makes AI integration prohibitively difficult. Building from scratch removes that dependency entirely, at the cost of significant upfront resource investment.
The customer-facing element of Meshed’s platform is a browser-based web application requiring no download. Clients log in with their email address and can track progress in real time. Whenever an update is received from an underwriter, it is processed through the company’s internal system, a proprietary broking tool built entirely by the team, and a notification is pushed to the customer portal with a summary and, where relevant, commentary from the Meshed team. Quotations, documentation, and policy comparisons are all visible in one place.
The internal AI system is named Rocky, “Tech and software people are notorious for bad naming,” he conceded, “but our AI system is named Rocky.”
One of the more distinctive features of Meshed’s proposition is what it describes as real-time cover adjustment. The mechanism is less dramatic than the marketing language might suggest, though the underlying capability is substantive. As part of the onboarding process, clients are encouraged to connect their accounting systems to the Meshed platform. That connection provides a continuous feed of financial transaction data.
“The detection bit is real time,” Liu explained. When a transaction is identified that falls outside the terms of an existing policy, or that suggests coverage limits may need revision, the system flags the issue immediately and the Meshed team initiates contact with the client to discuss adjustments. The practical effect is a monitoring capability that no traditional broking relationship, reliant on annual renewal cycles and ad hoc client contact, can easily replicate.
Data gathering also serves an underwriting purpose. Transaction patterns can reveal risk exposures that clients themselves may not have identified, and the depth of financial information available through accounting system integrations reduces the repetitive question-and-answer process that has long characterised the commercial renewal cycle.
It may appear counterintuitive that a company defined by its AI capability places customer experience, rather than technology, at the centre of its operational philosophy. Liu is clear that the two are not in tension, but that the sequencing matters. “Every broking workflow, every optimisation, automation, and AI application is anchored on providing the customer with a better experience,” he said.
The practical consequence of that orientation is that Meshed reversed the conventional broking process. Rather than completing the broking exercise and then presenting clients with a bundle of documents by email, the company designed its workflow around what the customer should be able to see at every stage. Insurers’ responses are automatically uploaded to the platform as they arrive, processed by AI, and made visible to the client in real time. The document-dump that traditionally marks the end of the renewal process is replaced by a transparent, continuous feed of progress.
Meshed’s approach to professional intermediaries departs from the conventional referral model in ways that reflect its broader philosophy. Where many brokers approach accountants and solicitors on purely commercial terms, offering a ten per cent referral commission in exchange for introductions, Meshed builds technology for its partners to use directly. That technology enables partners to assess clients’ insurance needs and cyber risk exposures, rather than simply pointing clients towards a broker.
The response has been notable. Liu reports that some partners decline the commission altogether: “They just want to help the customers,” he said. The accounting sector, already comfortable with platforms such as Xero and subject to government mandates around digital record-keeping, has proved particularly receptive to a technology-forward broking partner.
Cyber insurance forms part of Meshed’s product suite, and every onboarded client receives a complimentary cyber scan. The scan produces a score, identifies specific network configuration vulnerabilities, and provides remediation guidance. It also demonstrates, in concrete terms, the exposure a business faces without cyber cover in place. The offering is described as standardised cybersecurity assessment, positioned as a client service rather than a sales tool.
The rapid expansion of AI capability across financial services has brought renewed regulatory scrutiny to automated processes. Liu is candid about how Meshed manages that tension. “We always double-check the output from AI,” he said. “It’s always got a layer of human review.” The AI handles the volume of work: document scanning, data extraction, risk pattern identification, and notification. Human review remains a consistent requirement before anything consequential is communicated to a client or insurer. The model is designed to meet regulatory obligations without sacrificing the efficiency gains that make Meshed’s commission structure viable.
Meshed currently employs six people full-time. Two are on the technology team. The remaining four span broking, sales, and client operations. Liu describes the culture as “very lean and very flat,” with stand-ups held three to four times a week rather than daily, and a deliberate effort to give every member of the team genuine ownership of their work. “Every department is like its own little startup,” he said, capable of pursuing its own initiatives, setting its own priorities, and pursuing them independently, provided the work remains aligned with the company’s core objectives.
The company’s attitude to AI extends beyond the technology team. Every function, from sales to broking to engineering, is encouraged to experiment with AI tools in relation to its own workflows. Successful experiments are shared and, where appropriate, adopted company wide. Once a month, the entire team pauses operational work entirely for a hackathon: an open session devoted to exploring the latest AI tools, solving whatever problems seem interesting, and presenting findings to colleagues. The results feed directly into Meshed’s operational systems.
Asked to look three to five years ahead, Liu identifies the emergence of AI agents capable of functioning as genuine workforce members, rather than merely as software tools, as the development most likely to reshape the industry. He describes these as synthetic employees: on-demand agents with their own email addresses, personas, professional profiles, and telephone numbers, capable of conducting voice calls and managing relationships. “Hiring synthetic employees is probably the next thing,” he said, “not just for insurance broking, but other industries.”
Whether or not that framing proves prophetic, it reflects the direction of travel in AI development and the degree to which Meshed has positioned itself to benefit from it. The homegrown architecture was designed precisely to accommodate integrations that cannot yet be fully specified, because the technology is still, as Liu puts it, taking shape.
For all the intensity of building a regulated financial services business from scratch in under two years, Liu is resolute about maintaining distance from his screens at the weekend. He walks his dog regularly, cooks, and has taken up swimming with a characteristic commitment to long-term objectives. “My goal in five years’ time is to swim the English Channel,” he said, with the caveat that he only learned to swim two years ago. This summer, as an intermediate milestone, he will complete a five-mile open-water swim in the Lake District. His leisure reading, such as it is, leans towards science fiction; he recently finished the audiobook of Hail Mary by Andy Weir, and cites Ready Player One as another favourite. They are, perhaps, fitting choices for someone building a company on the assumption that the way the industry currently works will look, within a decade, like a relic of a much earlier age.
