Designers don’t want more AI; they want more control, writes Roderick Bates, head of product operations at Chaos
There is a belief that innovation offered by AI will automatically drive an increase in value. However, the more I speak with architects and visualisation professionals at every scale of practice, the more I believe this supposition falls flat.
Without a doubt, there is a strong demand for AI in the AEC sector, with our latest research showing 36% citing AI-assisted rendering as their top investment priority. Yet the same research reveals a tension: architects and designers’ top request is greater control over the AI they already use, pointing to a clear need for AI tools in AEC that can be directed, trusted and scaled within real design workflows.
Why architects don’t trust “black box” AI
The term “black box” AI typically refers to systems that take a basic input and produce an output, with little transparency or control on the process. For example, you put a prompt into an AI application and hope something greater comes out – but what happens in between is entirely hidden to you.
For the past few years, the dominant AI product experience has followed that pattern. For consumer applications, that experience is often fine. But for professional design practice, it isn’t.
Architecture is a discipline defined by intent. Every design decision, whether a material choice, a lighting condition or the way a space reads in a render, carries meaning and reflects the designer’s signature. When an AI tool generates an image that is visually compelling but disconnected from that intent, it creates an output that necessitates rework, confuses focused design intent and homogenises authorship.
In our recent research with nearly 800 architects and designers, the single most-cited obstacle to the application of AI in design was poor output quality and unreliable results, flagged by 48% of respondents – aligning perfectly with the number one improvement architects want to see in AI-powered visualisation tools: greater control over output images (33%).
This suggests architects are not resisting AI due to unfounded technophobia but rather from an unwillingness to surrender authorship to a system that doesn’t yet understand design intent.
The real demand is for controllable AI in architecture
Concept design and ideation are where architects are seeing the most value from AI, and it’s easy to see why. When speed and exploration matter most, AI definitely delivers. Almost nine out of 10 (86%) of current AI users report time savings, with nearly a third saying it has significantly sped up their workflow. At the concept stage, that figure rises to 48%.
From those early stages, the design process becomes less about design exploration and more about design definition. But with increasing project maturity, the need for precision increases and the tolerance for deviation decreases.
That value comes with caveats. Six in 10 architects and designers describe AI’s impact on their studio’s quality of work as “mixed”, enhancing some aspects of their workflow, while introducing new errors or inconsistencies.
The enthusiasm for AI is real and the time savings are proven but frustrations persist. What the industry is asking for is AI that doesn’t just generate but responds to and follows the designer’s intent.
That level of control would unlock AI’s value well beyond early exploration, carrying it into the later stages of design, where precision and intent matter most. This is the direction Chaos believes AI tools should evolve and one we’re actively building towards.
AI as a design companion
Architecture is not a field where outputs can be divorced from accountability. A render creates expectations and any misrepresentation of a material finish has consequences for client relationships, planning processes and construction.
The designer’s name is on the work, and so their professional reputation is implicated in every image that goes out the door. This is leading to growing interest in AI that lives within the design environment, which naturally aligns AI workflows with design intent.
When AI is integrated into the workflow, rather than bolted on to it, designers can apply their judgment in real time, steering their desired output, catching errors in context and maintaining the thread of intent from the original geometry through to the final image.
The distinction matters especially as AI outputs become more convincing. A technically brilliant image that misrepresents the design is more dangerous than a mediocre one, precisely because its realism makes it more likely to pass scrutiny.
A more considered relationship with AI
Our data describes an industry in the middle of a calibration process, learning the best applications for AI tools, where the current limitations lie and how tools need to evolve to achieve deeper integration into professional workflows.
The architects who are getting the most from AI are the ones who have made considered choices about where AI adds genuine value, maintained their professional judgment at every stage and pushed back when the technology asked them to accept outputs that fail to meet their standards.
The industry doesn’t need more encouragement to adopt AI, but it does need AI that fits seamlessly into the design process. Tools that respond to intent, support visualisation at every stage and can be steered with precision, while keeping creative control firmly in the hands of the designer.
That’s the value AI in AEC must deliver.
The post Giving designers the reins: The case for steerable AI in architecture appeared first on Planning, Building & Construction Today.
