Intelligence inspired by nature

Most AI decision-support tools are designed for a single decision-maker with a single objective. But real organisations are not like that. Consequential decisions span teams, functions, time horizons, and competing priorities. Centralising them creates bottlenecks — and loses the local knowledge that makes good decisions possible.

The assumption that one system can capture all relevant context, rank all trade-offs, and produce the right answer is both technically ambitious and organisationally counterproductive. It disempowers the people who understand the problem best.

Mycelia — the vegetative part of fungi — form vast underground networks that move nutrients, water, and chemical signals across ecosystems. These networks have no central processor, yet they solve highly complex resource allocation problems at scale, adapt to damage and disruption, and do so with extraordinary efficiency.

The key insight is that the network doesn't need to know everything centrally. Each node responds to local conditions; coordination emerges from the pattern of connections. Resilience comes from redundancy and adaptability, not from the strength of any single node.

This is the architecture we are building for human decision making.

Our methods treat decision processes as structured graphs — networks of interdependent choices, constraints, and information flows. We apply graph decomposition techniques to break large, complex problems into tractable subproblems that can be handled locally.

AI reasoning layers operate on this graph structure: propagating relevant context, identifying implicit dependencies, surfacing under-explored trade-offs, and flagging where additional information would most improve the quality of a decision.

Critically, the AI does not make decisions. It makes the structure of the decision visible — and ensures that the humans making it have what they need.

A common tension in organisational AI is between autonomy and coherence. Fully centralised systems impose coherence at the cost of local knowledge and agency. Fully decentralised systems preserve autonomy but fragment outcomes.

Mycelia resolves this tension by operating at the level of the decision network. Each node — a team, a department, an institution — retains full control over its own decisions and its own information. The AI methods coordinate at the boundaries: passing only what is necessary, surfacing only what is relevant, and ensuring that locally rational choices remain globally coherent.

This is not a compromise between decentralisation and coordination. It is both, simultaneously.

Mycelia does not assume that information flows freely. It is built for environments where sharing constraints are real — whether due to commercial sensitivity, regulatory requirements, competitive dynamics, or the simply practical reality that different teams track different things.

Each participant in the decision network contributes only what they can already quantify, in the form they already use. There is no requirement to adopt a common data model, centralise records, or expose information to parties who should not see it. The platform operates within these constraints as a first-class design requirement, not as an afterthought.

This makes Mycelia particularly well-suited to cross-institutional settings — consortia, partnerships, regulatory frameworks — where coordination is necessary but full transparency between parties is neither possible nor desirable.

The core algorithmic methods have been developed in collaboration with leading academic institutions across the UK. The framework has been validated on a broad range of benchmark problems spanning engineering, operations research, and network optimisation.

We are now applying these methods to the challenge of organisational decision-making — with a focus on practical usability, interpretability, and integration into the workflows that real teams already use.