What will we ask our buildings to be?

At a glance TL;DR

• Early AI demonstrations suggest that more building-production work could be delegated. This essay imagines a much more capable future, rather than claiming it has arrived.
• If design, technical knowledge and construction decisions stayed connected, qualities such as beauty, warmth and a relationship with nature might become easier to realise.
• That possibility leaves a human question: whose ambitions would guide the work, and how widely would its benefits be shared?

A floor plan, a short instruction, and an AI agent working inside Revit. In a September experiment, Sonu S George asked Astra to model a two-storey house from a supplied drawing, produce plans and elevations, and arrange them on a sheet. He reports a saved model and an exported drawing after about forty minutes. Some annotations overlapped and fixtures needed correction. It was a schematic starting point, with work still to do.

A week later, Autodesk demonstrated Claude changing window types across a model, duplicating views and applying room graphics through its expanded Revit connection. The connection uses MCP, a standard through which AI systems can use software tools. These previews put agents inside the working model, where their actions affect the information from which a building is documented.

For anyone accustomed to producing that information by hand, the significance is immediate. Work is being delegated through language. The practical questions concern what can be delegated, how dependable the results are, and what happens to the time those tasks once consumed. They also open a larger question: if intelligence becomes abundant throughout the process of making buildings, what kind of built environment could we ask it to produce?

To explore that, move forward in time. Today’s connections to Revit bring new capabilities into an existing arrangement of software, professions and deliverables. Imagine a later stage in which the process itself has developed around highly capable AI, in continuing conversation with the people who commission, design, make and inhabit buildings. This is a speculative destination. The demonstrations do not establish that it exists or tell us when it might arrive.

Consider it from the architect’s point of view. Their agent works with an evolving understanding of the building: what the client hopes to achieve, how the site behaves, how the design is developing and why particular choices matter. It develops alternatives and investigates their implications for structure, services, cost and construction. Where it needs knowledge or a decision from elsewhere, it consults the people involved or agents working on their behalf. It brings the consequences back into the conversation with architect and client, then carries the resulting decisions into the work.

That ability to develop and act across stages is the vertical integration envisaged here. Design, technical resolution and the means of building remain connected, so a discovery during procurement can inform the design as readily as a design decision informs procurement. How much capability sits within a practice, and how much comes from other organisations, remains open. The architect’s agent gives us a vantage point from which to understand the process; the eventual division of work could look very different.

Now consider what it would mean for this network to become dependable enough to carry a building project from intention through to occupation, with substantially less time and human production labour than it takes today.

Every architectural ambition has to be developed, tested, coordinated and made buildable. The time and skilled attention this requires place limits on what a project can attempt. If much of that work became readily available, ideas that would once have been too demanding to pursue could become practical. People with the desire and resources to build, but little capacity to navigate the process, could gain much greater agency within it. The capacity to turn an intention into a building could become far more widely available.

Consider the ambition to make a small hotel feel closely connected to its landscape. It reaches into where people arrive, what they see from a bed, how they move outside and where they find shelter. It concerns how the place feels in different seasons, alongside drainage, structure, servicing and the work of maintaining the grounds.

A sheltered courtyard brings these relationships into focus. Wind, sunlight, planting, circulation, overlooking and acoustic privacy all affect whether it becomes a place people want to spend time. Developing it involves discovering how those things interact.

The architect’s agent could explore that courtyard while drawing on the landscape designer’s knowledge and the operator’s experience of how guests use outdoor space. It could investigate wind and servicing implications, revise the design and bring alternatives back for discussion. Later, if a supplier suggested a different material, it could evaluate the change against the intended experience as well as cost and technical performance. Knowledge gained during construction could inform the next decision, with the reasons for earlier choices still available.

The project could also draw on experience beyond its own team. Where knowledge can be shared, a maintenance problem in another building might improve a detail; an unexpectedly successful shared space might suggest a possibility. Agents could help relate those experiences to the hotel’s particular circumstances. Each completed and occupied building could, in turn, contribute to the knowledge available for later work. The value of the network could grow through both its connections and what its participants learn.

In such a process, a relatively simple statement of ambition might have considerable reach:

Throughout the project, prioritise human experience: beauty, warmth, and connection to the surrounding nature of the site.

Its meaning would develop through the intelligence interpreting it and the people giving it substance. The agent might recognise implications nobody had articulated, propose a better arrangement, or find an economical way to achieve something previously out of reach. Architect, client and others could explore those possibilities together. What warmth or a convincing relationship with nature means in this particular building would become clearer through the work, finding expression in its materials, proportions, light, sound and use.

This possibility complicates the familiar fear of soulless AI architecture. Human authorship does not guarantee a humane building. Equally, machine involvement would not make a beautiful, comfortable place less valuable to the people using it. An excellent spatial idea remains valuable when an AI contributes it. The architect’s experience of making something and the occupant’s experience of inhabiting it both matter, but they are different experiences. Defending human creative work cannot, by itself, establish what produces the best places to live.

Nor is humanity in a building a single recognisable style. Warm colours, timber and planting can be welcome without establishing that the place responds well to its inhabitants. A hotel guest, a cleaner and a neighbour may experience the same design very differently. Taking human experience seriously would involve making those differences part of the conversation, including experiences less visible in a presentation image.

People and agents could help identify and reconcile those needs. Some apparent compromises might disappear as they find better ways to build. Others would remain choices about whose interests carry weight. Where the sheltered courtyard genuinely competes with another rentable room, a decision still has to be reached. The network’s ability to carry many voices would not give them equal influence. A commitment to human experience has consequence only if it continues to matter when other demands enter the project.

Architects have a substantial role in forming and sustaining those ambitions, opening them to scrutiny, and taking responsibility for what results. Clients, occupants and communities also bring knowledge and interests that should shape the work. Greater technical capability could make their participation more consequential by making more of their aspirations feasible.

The implications extend well beyond this imagined hotel. If exploring and resolving a particular response becomes inexpensive, the care available to a building could become less dependent on the size of its design budget. A modest project might benefit from an extent of investigation previously difficult to afford. Standard components could be assembled with greater sensitivity to particular lives and places.

Land, materials, energy and construction would still have costs. Lower design and coordination costs would not automatically deliver affordability, or ensure that savings reach occupants. But reducing the effort required to realise a thoughtful design could widen what is possible within an ordinary budget. That would be a significant change in who gets access to architectural quality.

The early demonstrations make this a question worth pursuing now, while the tools and working arrangements are still taking shape. Efficiency is valuable. So are the quieter achievements of a building: the room people return to, the threshold that makes them feel welcome, the place that becomes more meaningful through use.

If beauty, warmth and a meaningful relationship with place become much less costly to realise, how widely will we choose to make them available?

Sources

1. Sonu S George: I Gave AI a Floor Plan. It Built This in Revit, 22 September 2026 Retrieved 14 October 2026
2. Autodesk Developer Blog: Revit MCP server update, 29 September 2026 Retrieved 14 October 2026