The gap that determines everything
Every organisation has values. Very few have translated those values into the precise behavioural definitions that would make them operationally meaningful. The distance between a declared value and the behaviour that would consistently express it is not a communication problem. It is a design problem. And it is the gap in which most transformation fails.
CIT uses the term behavioural architecture to name the explicit design of the relationship between values and behaviour: the work of defining, with sufficient precision, what a value looks like in practice, how it is held across different levels and functions, how it is taught and governed, and how it is maintained under pressure. Without that architecture, values are aspirations. With it, they become the operating code of the organisation.
This research area has become significantly more important with the integration of AI into organisational systems. The behavioural architecture of an organisation does not govern only its human participants. It governs, or fails to govern, the AI systems operating within it. In a fused workforce, the behavioural architecture must extend consistently across both.
Four worked examples
The following four examples are drawn from CIT’s research and consultancy practice. Each illustrates the same underlying problem — a value that has been declared without being defined at the behavioural level — and the consequences of that failure in both human and fused workforce contexts.
Trust
Trust is among the most commonly cited organisational values. It is also among the most poorly defined. The distinction that determines everything is deceptively simple: is trust earned, or is trust given? If trust must be earned, then the default position of every participant in the system is one of withholding. They do not give their best, share their knowledge fully, or invest in collective outcomes until sufficient proof of the other party’s reliability has accumulated. The system operates on suspicion managed by evidence. Contribution is conditional. The culture that results is one of managed relationships rather than genuine collaboration. If trust is given — as a deliberate act, accompanied by explicit responsibility and accountability structures — then the default position becomes contribution. People can bring their full capability from the outset, because the cultural architecture tells them it is safe to do so. Neither position is declared by most organisations. Most organisations say trust is a value without specifying which behavioural model they are operating. The result is that different parts of the organisation are operating on different assumptions, some extending trust and feeling exploited, others withholding trust and appearing defensive. The value is present. The architecture is absent. In a fused workforce, this distinction determines how AI systems are configured and governed. An AI system built within a trust-is-earned culture will be deployed with extensive controls, limited autonomy, and constant human override requirements. An AI system built within a trust-is-given culture will be designed to operate with greater autonomy within clearly defined accountability boundaries. The outputs of those two systems, and the human experience of working alongside them, will be fundamentally different.
Knowledge
The conventional wisdom is that knowledge is power. That proposition has governed the behaviour of individuals in organisations for generations, and it is deeply embedded in the unconscious operating logic of most institutional cultures. If knowledge is your power, why would you share it? Sharing it means giving away what makes you valuable. The rational response, under that assumption, is to protect it. But organisations in the present period are attempting to build knowledge-sharing cultures. They invest in collaboration tools, knowledge management systems, communities of practice and internal communication platforms. And they consistently find that the knowledge does not flow in the ways they intended, because the behavioural architecture beneath the sharing aspiration is still operating on the power assumption. The alternative proposition is that knowledge compounds when shared. Under this assumption, sharing knowledge does not diminish its holder — it increases the collective intelligence available to the system, which generates new knowledge that the original holder can then access. The more knowledge flows, the more there is to share. The individual who shares generously becomes more valuable, not less, because they are connected to a network of returning knowledge rather than sitting on an isolated stock of it. These are not simply different attitudes. They produce different organisations. And in a fused workforce, the knowledge-as-power assumption is operationally destructive in a way it has never previously been, because the quality of AI output depends directly on the quality and generosity of human knowledge input. An organisation whose people withhold knowledge from each other will also, structurally, withhold it from the AI systems that need it to function. The assumption is the same. The cost is now compounded.
Purpose Versus Extraction
A third behavioural example operates at the organisational and investor level rather than the individual one, but the structure of the problem is identical. Most organisations declare a purpose beyond profit. Very few have defined the behavioural architecture that would make that purpose operationally real when it comes into conflict with short-term financial pressure. The distinction that matters is between value generation and value extraction. Value generation is durable: it creates something that did not previously exist, that can compound over time, and that strengthens the system from which it came. Value extraction is transient: it takes something from the system — the commitment of employees, the trust of customers, the resilience of communities — and converts it into a financial return without replacing what was taken. Organisations can do both. The question is which one their behavioural architecture actually rewards. In most organisations, the answer to that question is visible not in the values statement but in how performance is measured, how leaders are promoted, and what behaviour is tolerated when targets are under pressure. The purpose is declared. The behavioural architecture rewards something else. The gap between them is felt by everyone in the organisation, and it erodes trust, commitment and the quality of contribution at every level. This has direct implications for how AI is deployed. AI at scale will optimise for whatever the organisation’s behavioural architecture actually rewards. If the architecture rewards extraction, AI will extract more efficiently. If it rewards generation, AI will generate more effectively. The technology amplifies the existing behavioural logic. It does not change it.
The AI Measurement Gap
The fourth example is not drawn from the behaviour of individuals or organisations. It is drawn from the behaviour of AI systems themselves — and from the gap between what organisations believe they are building and what they are actually building. Most organisations deploying AI describe their intention in terms of augmentation: they are building systems that will help their people work better, think more clearly, and contribute more effectively. They are, in their own framing, building toward an OAI model: human original intelligence at the centre, AI amplifying it. But when CIT examines the behavioural architecture of those deployments — the actual parameters governing how the AI system operates, what it optimises for, how it is configured in relation to human judgement, and what the human experience of working alongside it actually is — the picture is consistently different. The AI has been designed to structure workflows, reduce variance in human decision-making, increase processing speed, and manage output quality within defined tolerances. These are not OAI design objectives. They are AI-in-the-loop design objectives. The human is a participant in a system the AI has structured. The gap between the declared intention and the operational reality is not usually the result of bad faith. It is the result of the absence of a behavioural architecture for AI deployment: a precise definition of what augmentation means in practice, what the boundaries of AI autonomy are, where human original intelligence is protected and where it is substituted, and how the alignment between human and AI behaviour is measured and governed. Measuring that gap — between the AI behaviour an organisation intends and the AI behaviour it has actually built — is one of the most important things CIT now does. It is the AI equivalent of the values-to-behaviour gap in human culture. And it is, in most organisations, substantially larger than anyone has yet acknowledged.
The design implication
Behavioural architecture is not a soft capability. It is a governance discipline. The organisations that will deploy technology most effectively in the coming decade are those that have done the work of defining, with precision, what their values mean at the level of specific, observable, governable behaviour — and have then extended that definition consistently across human and digital participants alike.
That work does not happen by itself. It requires structured research, honest diagnosis, and the willingness to examine the gap between aspiration and operational reality without defending the gap. CIT’s research in this area provides both the framework and the instruments for that examination.
