somai

somai is a studio that builds AI-native organisations. We partner with established companies seeking to transform their core operations through AI, rather than applying it only at the margins.

The challenge we address is widespread. Many organisations have adopted copilots, provided staff with access to foundation models, and run pilots. While these efforts are often useful, they tend to be generic because the underlying operating model remains unchanged. Investments rarely impact the profit and loss account. The issue is structural, not technical: the tools function, but are added to companies not designed for them.

We take a foundational approach. To become AI-native, a company needs core systems in place before AI can deliver meaningful results: a central knowledge base, a trusted data foundation, a decision architecture, a communication framework, an operations backbone, external sensing capabilities, and a compliance core. Most organisations have not built these deliberately. We implement them sequentially, tailored to each business.

We start with a blueprint: a written brief and a value hypothesis agreed by both parties. Our senior team works directly with the future operators to deliver a production-ready solution. Upon handover, the client receives the code, data, and an operator manual.

Our commercial model reflects this approach. Fees are fixed based on a value case, not billed by the day, with a portion at risk if the agreed value is not delivered. Clients own all deliverables, ensuring no ongoing dependency. We prefer to be evaluated on long-term capability rather than engagement size.

Somai began with an observation independently made by three people from distinct perspectives.

One founder spent two decades building a regulated data business, expanding to over forty-five countries through a single integration and understanding the challenges of ensuring reliable information at scale. Another spent thirteen years at a leading technology company, developing two global businesses from concept to profitability in a highly collaborative environment. The third spent twenty years in tier-one banking, establishing fraud analytics and compliance functions subject to regulatory scrutiny.

Despite different backgrounds, we observed the same trend: companies invested in AI but saw limited returns. The issue was not the technology, which is effective, but its integration into organizations designed for a pre-AI world. An MIT study found that about ninety-five percent of enterprise AI pilots never impact the profit and loss account. This is often seen as a technology issue, but we believe it is a context issue. The models are capable, but they are deployed in businesses lacking the necessary visibility.

We concluded that being AI-native is a structural, not generational, characteristic. A recent start-up may not be AI-native, while an established company can be, depending on whether its operating model assumes intelligence is abundant rather than when it was founded.

This conclusion led to an important realization that defines Somai’s purpose. If the problem is structural, advice alone is insufficient. The foundations must be built: a reliable record of company knowledge, trusted data, effective decision-making processes, and robust controls. For this reason, we did not start a consultancy.

Instead, we established a studio: a team that builds the solution, transfers ownership, and ensures the client retains full control.

James Varga

FOUNDING PARTNER

EDINBURGH, LONDON HYBRID

linkedin.com/in/jvarga

Leads somai’s commercial work, and owns the studio’s own company brain — the same foundation somai installs for clients, built on somai first.

Previously founded and scaled a regulated data business operating across more than forty-five countries, holding a US patent on the underlying approach. somai’s position on trustworthy data inside regulated environments comes from having built it, not advised on it.

Matthew Bilsland

FOUNDING PARTNER

EDINBURGH

linkedin.com/in/matthewbilsland

Leads delivery and commercial development, and the studio’s published thinking.

somai’s clients are established organisations carrying real institutional complexity, not start-ups. Matt spent thirteen years building new global businesses inside exactly that kind of environment — which is where the studio’s method comes from: build in production, alongside the operators who will run it, with no parallel shadow systems.

Swagatam Sen

FOUNDING PARTNER

UNITED KINGDOM

linkedin.com/in/swagatam-sen-data-science

Leads the technical and regulatory work, including somai’s pattern-detection architecture.

Twenty years building fraud analytics and compliance functions inside tier-one banking, including group-wide systems required by financial regulators. That is the basis for somai’s position that governance is an engineering discipline rather than a documentation exercise — and that AI has to reach production under genuine regulatory exposure to count.