Businesses spend a remarkable amount of time maintaining order.
Information is collected, organised, reviewed, reconciled and communicated. Meetings establish a common understanding, reports describe progress, managers coordinate responsibilities and processes ensure that people work towards the same objectives.
Much of this activity can be understood as an effort to resist a particular form of organisational disorder: information entropy.
The interesting possibility is that AI could fundamentally change how much effort organisations need to expend maintaining that order.
Understanding information entropy
In thermodynamics, entropy describes a fundamental property of physical systems associated with the number of microscopic arrangements consistent with their overall state. In information theory, Shannon entropy measures uncertainty in information.
Neither definition directly describes organisational inefficiency. But entropy provides a useful analogy for what happens to information and shared understanding within a business.
As organisations grow and activities become more complex, information spreads across people, documents, applications and departments. Different interpretations emerge, knowledge becomes disconnected from its origins and the organisation's understanding of its own activities gradually diverges.
Maintaining coherence requires continuous effort.
The problem is not simply that information becomes disorganised. It is that different parts of the organisation increasingly operate with different understandings of the same reality.
How entropy develops
Consider a complex technology programme.
A supplier reports a delay. The project manager updates a schedule, a commercial manager considers the contractual implications, a technical lead assesses the affected interfaces and a programme manager records a new risk.
Each person may perform their responsibilities competently, yet the information now exists in several places, expressed differently and updated at different times.
A subsequent change might reach some of these records but not others.
The schedule may reflect a revised completion date while the risk register retains the previous assumptions. A management report may describe an issue as resolved while the technical team is still working on it.
No single failure caused this divergence. It emerged through the ordinary operation of the organisation.
The same effect occurs when decisions lose their rationale, responsibilities become unclear, knowledge remains with individuals or successive handovers gradually alter the interpretation of requirements.
As complexity increases, maintaining a consistent picture of reality becomes progressively more difficult.
The cost is ultimately predictability
The visible symptoms of information entropy are familiar: duplicated work, conflicting reports, repeated meetings, unresolved questions and time spent searching for information.
But the more significant consequence is a loss of organisational predictability.
Decisions are made using incomplete or outdated information. Dependencies are overlooked. Problems are recognised late. Commitments are made without a reliable understanding of the work required to fulfil them.
The organisation becomes less capable of accurately predicting cost, schedule and outcomes.
Information entropy therefore has consequences beyond administrative efficiency. It affects the organisation's ability to coordinate activities and deliver its objectives.
How businesses traditionally maintain order
Organisations have developed extensive mechanisms to counter these effects.
Databases, enterprise applications and document management systems provide controlled repositories for information. Defined processes and responsibilities establish how work should be performed. Governance arrangements introduce oversight, while reports, meetings and reviews help reconcile different perspectives.
Project Management Offices are a particularly visible example. Schedules, registers, forecasts and reporting cycles are all intended to maintain a consistent understanding of delivery.
These mechanisms are necessary and often effective, but they introduce their own costs.
Information must be entered, transferred, checked, updated and reconciled. Different management artefacts become additional representations of the same underlying events, each requiring maintenance.
The systems intended to control information entropy can themselves become sources of it.
The limits of periodic reconciliation
A recurring feature of traditional management is that order is restored periodically.
Schedules are updated weekly. Risks are reviewed monthly. Reports are prepared against agreed reporting dates. Teams meet to resolve differences in their understanding.
Between these activities, reality continues to change.
The resulting management information is therefore often a reconstruction of what has happened, assembled after the events themselves.
Increasing the frequency of reporting can reduce this delay, but usually requires additional administrative effort.
The underlying problem remains: the organisation depends on people repeatedly reconstructing a shared understanding from information distributed across multiple sources.
AI as a continuous mechanism for maintaining coherence
AI introduces a different possibility.
Rather than waiting for information to be manually collected and reconciled, AI could interpret incoming information as it becomes available.
Supplier communications, progress observations, technical records, approvals, contractual changes and operational events could be associated with the relevant entities, activities and dependencies.
An AI-enriched information system could recognise that a supplier's revised delivery date affects an installation activity, that the installation is a prerequisite for testing, and that the change may place a contractual milestone at risk.
It could identify inconsistencies between records, preserve the reasoning behind decisions and bring unresolved questions to the attention of the appropriate people.
The underlying records would remain authoritative. AI would maintain a richer representation of their relationships, implications and uncertainties.
Reports, risk assessments, forecasts and management views could then be generated from this shared information rather than maintained as separate, independently updated descriptions of reality.
The distinction is important.
Instead of periodically restoring a common understanding, the system could continuously maintain one.
This would not eliminate the need for management. It could, however, substantially reduce the effort devoted to collecting information, reconciling contradictions and reconstructing context.
AI can also increase entropy
There is an important risk in this approach.
AI can misinterpret information, infer relationships that do not exist, generate convincing but incorrect explanations and propagate errors between systems.
If these interpretations are accepted as facts, AI could accelerate the very information disorder it is intended to reduce.
The architecture therefore matters.
Original evidence must remain accessible. Authoritative records must be distinguished from AI-generated interpretations. Uncertainty, conflicting information and the history of decisions must be preserved.
Changes to contractual obligations, financial commitments, accepted scope and other consequential records require appropriate validation and authority.
A continuously maintained semantic information layer could support these controls by linking interpretations to their sources and recording how the organisation's understanding changes over time.
The objective is not for AI to independently decide what is true. It is to help maintain a reliable, explainable representation of what is known, what is uncertain and what requires a decision.
From information management to organisational coherence
The implications extend beyond better reporting or more efficient document management.
Much of management exists to keep people, information and activities aligned with organisational objectives.
Structure establishes responsibilities. Processes coordinate work. Competency enables sound judgement. Culture influences how information is shared and problems are addressed.
AI could provide a further mechanism for maintaining that alignment by making relevant information, relationships and decisions continuously available.
Managers could devote less effort to assembling a shared understanding and more to making decisions, resolving problems and directing work.
The organisation would still generate new information, face uncertainty and experience disagreement. Those are inevitable characteristics of complex human activity.
But maintaining coherence could become a normal capability of the information system rather than an activity requiring repeated manual intervention.
Traditional management fights information entropy through continuous human effort, often organised around periodic reconciliation. AI creates the possibility of maintaining organisational coherence continuously, with far less administrative work.
That may prove to be one of AI's most significant contributions to how businesses operate.