Ivory & Co.
AEA by Ivory & Co.StandardsKnowledge & ResearchImplementationCertificationIvory & Co.

Models in AEA by Ivory & Co.A model is a way of seeing. What it reveals depends on what it leaves out.

A model selects some of what a business knows about itself and presents it for a purpose. The business, the knowledge about it and the picture of it are three different things.

A CEO opens the monthly performance dashboard. Revenue is up. Margins are stable. Customer satisfaction is improving.

Across the table, the operations director has a different view. Delivery delays are increasing. Employees are working overtime. Several important customers have begun raising concerns about reliability.

The CEO looks back at the dashboard. Everything is green.

Which picture is correct? Possibly both. The dashboard may accurately show the measures it was built to display, and the operations director may be describing conditions those measures never captured. Neither gives a complete understanding of the enterprise.

This is the central challenge of modelling. We build representations to make complexity understandable. We select what matters for a purpose, organise it and present it in a form people can interpret. That makes models extraordinarily useful, and it makes it easy to mistake the representation for the thing represented.

The enterprise is larger than any picture of it

Ask five specialists to describe the same business. The strategist produces a map of markets, competitive position and priorities. The finance director presents economic performance, capital requirements and financial risk. The operations director describes the work, the resources and the constraints on delivery. The technology architect draws systems, applications and information flows. The people director presents the organisation, its workforce and their capabilities.

Each representation may be accurate, and each reveals something different. All five specialists are looking at the right enterprise, each through one lens. The business contains every one of those concerns, their relationships and a good deal more.

A model’s usefulness depends on the question it was built to answer. A financial model can be excellent on cash flow and silent on how customer enquiries are handled. A process diagram shows how work moves between activities without establishing whether the business has the capabilities to perform them. An organisational chart shows reporting lines and leaves out many of the dependencies the enterprise runs on.

Those are legitimate differences in purpose. Problems start when a model is asked a question it was never designed to address.

Three things worth keeping distinct

A retailer wants to understand why customer returns are increasing.

There is the business itself: customers buying products, deliveries being completed, products coming back, employees handling enquiries, refunds being processed.

There is what the business knows about those activities: transaction records, customer feedback, delivery information, operating instructions and what employees have observed.

And there are the representations built from that knowledge: dashboards, reports, process maps, forecasts and analytical models.

The three are related and none can stand in for another. The enterprise keeps operating whether or not anyone has described it accurately. Knowledge about it may be incomplete, outdated, contradictory or uncertain. A model selects and organises some of that knowledge for a purpose.

The difference becomes commercially significant the moment a model influences a decision. Suppose the returns dashboard shows that most returns occur within seven days of delivery. Leadership may conclude that customers are dissatisfied with their purchases. That may be true. The same pattern could come from damaged products, incorrect fulfilment, misleading descriptions, or customers ordering several sizes intending to send some back.

The dashboard describes a pattern. The cause has to be found in the enterprise and the evidence. A representation can be accurate and still be insufficient.

Why businesses become attached to their models

Models give us something complex organisations rarely offer naturally: a sense of order. An organisational chart turns hundreds of relationships into a manageable structure. A strategy map gives direction a visible form. A dashboard reduces thousands of transactions to a handful of indicators. A process diagram makes work look sequential.

These representations are valuable because they simplify, and simplicity has a psychological effect. Once something is visible and familiar, we begin to treat it as the thing itself. A department becomes synonymous with the capability it delivers. A job title becomes the work. A system becomes the business function it supports. An indicator becomes the outcome it was meant to measure.

Consider a business that has always organised customer service, sales and marketing as separate departments. Its leadership team sets out to improve the customer experience, and the discussion quickly turns into a negotiation between the three. Who owns the customer journey? Who leads the project? Which department gets the investment?

Those questions may need answers, but the organisational model has already shaped the conversation. Customer relationships and service requirements do not divide neatly along departmental lines. The model is useful for understanding the organisation. It is a poor starting point for redesigning the customer experience, and recognising that makes room for a different examination.

Enterprise modelling: different questions require different representations

Suppose the retailer investigates its returns problem properly. It may need several models. A customer journey shows what customers experience from purchase through delivery to return. A process model explains how returns are received, assessed and resolved. A financial model examines the commercial effect of different return rates and handling costs. A capability view establishes what the business needs to be able to do. An operational dashboard shows current performance. A scenario model explores what might follow from changing the policy.

Each serves a different purpose, and together they reveal relationships that are hard to see separately. The customer journey shows that customers receive inconsistent return instructions. The process model shows that different sales channels follow different procedures. The financial analysis shows that delays in resolving returns are raising handling costs.

Combining representations leaves their distinctions intact. A customer experience remains a different subject from a financial forecast. A process description is different from evidence that the process is being followed. A proposed future is different from the current operating condition. The value comes from seeing how the perspectives relate while preserving what each one means.

The danger of competing truths

Two departments maintain separate versions of the same operational process. Operations has a detailed diagram. Customer service has a simplified one. Both are used for training.

Over time the diagrams diverge. One includes an approval step the other omits, and employees follow different instructions depending on which document they were given.

It is tempting to solve this by declaring one diagram official. That may be necessary. The deeper question is what knowledge each diagram represents. If both describe the same governing requirements, both should agree with those requirements. If they describe different circumstances, the difference should be explicit. If one is outdated, the organisation needs to know what changed.

An enterprise that keeps the knowledge separate from the forms in which people consume it can provide different views for different audiences without contradiction. A manager needs an overview. An employee needs detailed instructions. An auditor needs evidence of the conditions under which work was done. The representations differ, and all of them rest on the same coherent knowledge.

When the model becomes the design

A company plans to automate its purchasing function. Its first step is to map the existing process. The diagram shows a purchasing officer receiving a request, checking stock, contacting suppliers, comparing quotations and preparing a purchase order. The automation team builds a solution that follows those steps. The result is faster.

Has the enterprise been redesigned? Perhaps the purchasing officer’s steps reflected the limits of earlier systems. Perhaps supplier information could be available continuously. Perhaps replenishment could be driven by demand forecasts, with no individual request at all. Perhaps certain approvals are only needed under particular commercial conditions.

The existing process model describes one way the work has been arranged. When it becomes the unquestioned design, transformation reproduces historical constraints in new technology.

This is especially relevant to AI. An enterprise can spend substantial resources teaching agents to imitate its existing workflows, and gain useful efficiencies. The larger opportunity may lie in reconsidering the requirements and designing a different way to meet them. Models should help us examine those possibilities, and a good one never quietly closes them off.

Models of the present and models of the future

A business often needs to understand several conditions at once. What is operating today? What has been approved and not yet implemented? What could the business look like under an alternative? What might follow from adopting it?

These produce representations that look remarkably alike. A current-state process diagram and a future-state one use the same symbols. A forecast dashboard resembles an actual performance dashboard. A proposed organisational chart is identical in format to the current one.

Their meanings are different. One describes an established or observed condition, another an intended arrangement, another a possibility. The format says nothing about the status of the knowledge inside it.

So the context of a model matters. A reader needs its purpose, scope, source, version and limitations, and needs to know whether it shows something established, proposed, observed or uncertain. A compelling picture of a future enterprise is still a picture of something that may not yet exist.

When models disagree

Return to the CEO’s dashboard. Revenue is rising while operational pressure builds. The immediate temptation is to decide which view is correct. It is more useful to examine what each one establishes.

The dashboard shows that the business is generating more revenue. The operations director shows that the resources and arrangements supporting that growth are approaching their limits. Both are true, and together they reveal a commercial risk that neither communicates alone.

This is one of the most valuable uses of models. Disagreement between them can expose incomplete information, inconsistent assumptions, or a real difference between what was designed and what is happening. The disagreement has to be investigated. No model earns authority by being more sophisticated, better looking or more often used.

AI will make representations almost effortless

The cost of producing business models is collapsing. AI already helps generate diagrams, summarise processes, prepare reports, analyse data and build scenarios. Before long it may be possible to ask a question about an enterprise and receive a tailored representation almost immediately. Show me how this customer promise is fulfilled. What would change if we outsourced distribution? Which activities depend on this supplier?

These are useful questions. The quality of the answer depends on the knowledge available and the integrity of the relationships it draws on. An AI-generated process diagram can be beautifully presented and omit an important exception. A forecast can contain precise numbers built on uncertain assumptions. A model can combine information from different periods without recognising that the business has changed in between.

Speed solves none of this, and can make it harder to notice. As pictures become easy to produce, the challenge shifts to knowing what a picture legitimately tells us.

A model should help us see further

The dashboard is green and the operation is under pressure. The business has no need to abandon the dashboard. It needs to understand what the dashboard shows, what it excludes and how its measures relate to the actual condition of the enterprise.

Perhaps it should include indicators of operational capacity. Perhaps leadership needs a different representation for emerging risks. Perhaps the existing measures reward outcomes that create difficulty elsewhere. Each possibility deserves examination, and the essential thing is that the model stays open to challenge.

In Adaptive Enterprise Architecture, a model is an instrument of understanding. The same knowledge can be expressed differently for different audiences, and no particular diagram, document or software format is treated as the definitive description of the business.

The purpose of a model is to help us see the enterprise more clearly. The discipline lies in remembering that there is always more to see.

The Models Library

The Models Library will hold approved representations that support the understanding and application of Adaptive Enterprise Architecture. Each published model states its purpose, scope, source, version and limitations, so that readers know what it represents and how to use it.

The library opens as individual models are approved for public release. None is published yet.

Not yet publishedSource diagrams, structural schemas and modelling conventions.

Keep readingWords are part of the architecture.Standards · Glossary →
→Continue

Where to go next.