Institutions often fail to use knowledge they already possess, not because the knowledge is missing, but because it exists in a form that other parts of the institution cannot readily understand or apply. A legal team may describe a problem in terms of statutory authority and procedural risk, while an economist frames the same issue through incentives and behavioural effects, an engineer through system constraints and an operational unit through cases, exceptions and workload. Each perspective may be valid. Yet validity alone does not make knowledge institutionally usable.
This is one of the less obvious consequences of specialisation. Modern institutions depend on people who know different things deeply, and that differentiation is a major source of capability. Expertise allows complex problems to be analysed with precision that general knowledge could never achieve. But the very processes that produce sophisticated expertise also produce specialised vocabularies, conceptual distinctions, professional assumptions and standards of evidence. As knowledge becomes more advanced inside each domain, it can become harder to move across the boundaries between them.
The difficulty is sometimes described as a communication problem, but that description is too shallow. Communication concerns whether information has been transmitted. Translation concerns whether meaning survives the movement. A report can be circulated perfectly and still fail to become usable knowledge elsewhere in the institution because the people receiving it do not share the interpretive framework through which it was produced. The words arrive; the cognitive structure required to understand their significance does not.
Imagine a technical team warning that an infrastructure programme contains an unacceptable level of systemic risk. The phrase may carry a precise meaning within engineering practice, tied to particular assumptions about failure, redundancy and tolerance. A finance team may read the same language through a different framework, interpreting “risk” primarily as probability multiplied by expected cost. Senior decision-makers may hear a more generic warning that something could go wrong. Everyone has received the same words, but they may not have received the same knowledge.
This problem becomes more difficult because institutions often assume that shared vocabulary implies shared meaning. Terms such as resilience, efficiency, vulnerability, capacity, risk, performance, evidence and impact circulate widely across public organisations. Their familiarity can create an illusion of semantic alignment. Yet the same term may encode very different distinctions in different professional communities. A statistician, lawyer, social worker and cybersecurity specialist can all speak about “risk” while referring to substantially different objects.
Semantic incompatibility does not therefore require different languages in the ordinary sense. It can occur inside a single language and even inside a single organisation. Two departments can use identical words while organising reality through different conceptual structures. Translation becomes necessary not only when terminology differs, but when the assumptions, categories and relationships behind the terminology differ.
Institutional translation is the capability to make those structures mutually intelligible enough for knowledge to travel without being stripped of what makes it useful. This does not mean turning every specialist account into a simplified summary. Simplification can sometimes help, but excessive simplification can remove precisely the distinctions that give expertise its value. Translation must preserve enough of the original knowledge to make it faithful while reconstructing it in forms that another community can use.
That reconstruction is demanding because some concepts do not have direct equivalents across domains. A legal judgement about proportionality cannot simply be converted into a numerical variable without losing important meaning. An ethnographic account of community behaviour may resist the categories used in administrative statistics. A machine-learning model may identify relationships that operational staff cannot easily connect to existing procedural knowledge. Translation sometimes requires building a bridge rather than finding a synonym.
Good translation therefore begins by asking what another part of the institution needs to understand in order to act intelligently. The answer may not be every technical detail. A minister does not need to become an engineer to understand the implications of an engineering assessment, and an engineer does not need to become a constitutional lawyer to understand a legal constraint. What matters is that the receiving actor acquires the distinctions necessary to understand what can be done, what cannot, what remains uncertain and what would be lost by treating the knowledge too crudely.
This is why translation should not be confused with standardisation. Standardisation seeks common forms, definitions or procedures so that different parts of an institution can operate consistently. That can be extremely useful. But not every difference should be eliminated, and some forms of expertise cannot be reduced to a single common vocabulary without becoming poorer. Translation allows distinct knowledge systems to remain distinct while still becoming interoperable.
The difference resembles the distinction between requiring everyone to think in the same language and developing the ability to work across languages. The first approach reduces diversity in order to reduce friction. The second attempts to preserve diversity while making coordination possible. Institutions often need both, but they solve different problems. Standardisation can remove unnecessary variation; translation allows necessary variation to coexist with shared action.
Translation also changes the character of institutional decision-making. When specialist knowledge cannot cross boundaries, decisions tend to rely disproportionately on whichever forms of knowledge are already legible to the centre of the organisation. Quantitative indicators may dominate because they travel easily. Legal formulations may dominate because they map directly onto formal authority. Operational experience may be undervalued because it remains embedded in narratives and tacit judgements. The result is not necessarily that the institution lacks relevant knowledge, but that some knowledge is easier to transport than others.
Over time, this can produce systematic cognitive bias. Forms of expertise that translate easily into common institutional formats gain influence, while those requiring more interpretive work may appear secondary, subjective or difficult to use. An institution can therefore become epistemically unbalanced even when every specialist community is individually competent. The problem lies in the architecture connecting them.
Translation capability helps correct this imbalance by making the movement of knowledge an explicit institutional task. That task may be performed by boundary roles, interdisciplinary teams, shared interpretive practices, joint problem framing or people who understand enough of multiple domains to mediate between them. The organisational form can vary. The underlying function is the same: ensuring that knowledge produced in one cognitive environment can become meaningful in another.
The need becomes even clearer in situations where decisions cross organisational boundaries. Climate adaptation, public health, digital regulation, infrastructure planning and AI governance all require knowledge from multiple professional domains. A technically correct answer inside one discipline may still be institutionally inadequate if it cannot be related to legal authority, political feasibility, operational capacity, public behaviour and competing forms of evidence. Complex governance therefore depends not only on accumulating expertise but on making expertise cognitively interoperable.
Artificial intelligence adds another translation boundary. AI systems can produce outputs in forms that are statistically sophisticated but difficult for decision-makers to interpret, while institutional users may pose questions through categories that poorly reflect how models operate. Translating between model outputs, professional judgement, administrative categories and public reasoning becomes part of the institutional task. The problem is not merely explaining AI to non-experts. It is making machine-generated knowledge compatible with the forms of reasoning through which institutional decisions acquire meaning and legitimacy.
Translation should not be understood as a demand that disagreement disappear. Two professional communities may fully understand one another and still disagree because they value different outcomes, accept different risks or interpret evidence differently. Translation does not resolve substantive conflict. It makes the conflict more intelligible by ensuring that disagreement is not merely the accidental consequence of mutual incomprehension.
This distinction matters because institutions can otherwise mistake semantic friction for substantive disagreement. One department may appear resistant to another’s proposal when it is actually responding to a different interpretation of the same term. Conversely, superficial agreement may conceal semantic divergence when participants use the same vocabulary but attach different meanings to it. Translation helps reveal what the disagreement is actually about.
A cognitively capable institution therefore does more than distribute information. It develops ways for knowledge to cross professional, organisational and technical boundaries without losing its meaning. This capability is especially important when the institution already has the right knowledge somewhere inside it but cannot connect that knowledge to the place where a decision is being made.
The institutional question is no longer simply, “Do we know this?” It becomes, “Can the knowledge we possess become usable by the people who need to act on it?” The difference is fundamental. Knowledge that remains trapped inside a specialist language may be intellectually valid yet operationally inert.
Translation is part of institutional intelligence because institutions do not think through a single vocabulary, discipline or professional worldview. They think through multiple forms of expertise whose meanings do not automatically align. The challenge is not to eliminate those differences, but to make knowledge capable of moving across them without being distorted beyond usefulness. An institution can possess the right knowledge and still fail to act intelligently if that knowledge exists in the wrong language for the people who need it. Institutional intelligence therefore depends not only on what the institution knows, but on whether its different ways of knowing can understand one another.