How AI could keep a design project connected

At a glance TL;DR

• A design change affects more than the model. Its reasons and consequences also need to carry through drawings, schedules, documents and conversations.
• This essay imagines AI keeping that project context connected, so people spend less time reconstructing decisions and reconciling separate accounts.
• Useful coordination would preserve uncertainty and disagreement, and distinguish a proposed change from an agreed decision. This is a possible direction for practice, not a demonstrated system.

A client asks to make a meeting room larger. The partition moves in the model, but the decision also has a life in the furniture layout, room schedule, specification, cost plan and the email that explains why it was made. Someone has to work out what follows, carry the change through and check that the accounts still agree.

AI could make each of those operations quicker. The larger possibility is that it could keep them connected: remember the reason for a change, follow its consequences and carry the agreed decision through the work. That would change what people spend their time coordinating, and potentially how a practice organises itself.

This essay imagines that possibility inside a design practice. Its subject is the work between a decision and the many places where that decision must take effect.

A decision has more than one form

A project needs different accounts of the same building. A plan shows relationships that a schedule cannot. A specification sets out requirements that geometry alone cannot express. A conversation may explain why an apparently awkward arrangement matters to the client.

Those differences serve a purpose. The difficulty is keeping their relationships intact as the project changes.

The larger meeting room might reverse an earlier decision to give more space to a neighbouring team. Its new furniture arrangement might depend on equipment that has not been selected. The client’s email might ask for an option to be explored, without authorising a change to the current design. Updating every document immediately could produce a beautifully consistent mistake.

An integrated system would therefore need to work with reasons, alternatives and the status of decisions as well as dimensions and quantities. It would have to preserve what remains unresolved.

The project becomes the place of work

Imagine opening the project and asking what would follow from enlarging that room. The system brings together the current layout, the earlier decision, the client’s request and the relevant requirements. It prepares an option and identifies the consequences it can establish, alongside questions it cannot settle.

The team can inspect the drawing, alter it directly and compare the result with the earlier arrangement. Where the proposal depends on another person’s knowledge, the question goes to them with enough context to answer it. Their response becomes part of the same decision, available wherever it matters later.

Once the change is agreed, the system carries it through the affected material and shows what it has changed. If two accounts conflict, it makes the conflict visible. Yesterday’s issued drawings remain yesterday’s record; a new option does not quietly rewrite them.

In this imagined practice, the evolving project becomes the main place of work. Models, documents and specialist applications provide different ways to examine and develop it. People could move between them without repeatedly assembling the context themselves.

That would be a more substantial change than adding a separate assistant to every application. A capable assistant can still leave its user with the job of remembering what every other assistant needs to know. The promise here is continuity across the work.

Coordination changes its shape

If that continuity became dependable, a practice might spend much less time reconstructing what has happened. A colleague returning to a project could see the decisions since their last involvement, why they were made and what remains open. A review could begin with the actual points of disagreement, with less preparation spent collecting and reconciling the documents.

The system might also recognise consequences that nobody had asked it to look for. A room change could reopen a previously settled question elsewhere in the plan. It could bring that relationship back into the discussion while there was still time to consider it.

This could loosen an organisation built around people tending separate applications and deliverables. A small team might carry a project through more of its development, drawing on specialist knowledge when the work calls for it. Software expertise would still matter, but less of the practice’s coordination would depend on individuals remembering the routes between its tools.

There is a danger in making the project appear more settled than it is. The client and designer may mean different things by the same request. Two consultants may disagree for good reasons. An older decision may deserve to be challenged. A system that smooths these differences away would make the work harder to understand, even if its documents matched perfectly.

Keeping the project connected would have to include keeping disagreement legible. People would need to see how an answer was reached, question it and return to an earlier state. Their understanding of the project would remain essential to its development.

What disappears from the working day?

Some exchanges bring another person’s judgment into the work. Other transfers are needed simply because the information sits in separate applications. The opportunity is to reduce the effort of crossing those software boundaries while making the exchanges between people better informed.

The meeting-room decision would still require discussion. Its consequences would still require judgment. But the team could spend less of its day moving the decision from place to place, checking whether it arrived and explaining its history again.

An architectural practice organised around that continuity could feel quite different from one organised around a succession of files. More of the working day could be spent developing the project, with its reasons and unresolved questions still in view.

Corrections

1. 14 October 2026 — Retitled and substantially revised to focus on how AI could connect project information and decisions within a design practice.