With foundation models unlocking new potential across the built environment, architecture and construction are finally ready for AI. However, instead of replacing creativity, Nemetschek’s strategy is to augment it, focusing on ethical deployment, real-world workflows, and accelerating access to domain expertise.
Digital technology has promised to transform the architecture, engineering, construction, and operations (AECO) industry for decades. Yet despite the widespread adoption of CAD, BIM and cloud collaboration tools, productivity has stagnated. Each building remains a unique prototype, crafted in bespoke environments by dispersed teams juggling non-standardised data. Compared to manufacturing or finance, AECO has always lagged in AI adoption, not because it lacks imagination but because of the consistency and structure that conventional AI used to require.
Julian Geiger, VP, Head of AI at Nemetschek, believes that is finally changing. “Before, what I would now call classic AI required extensive data and highly specific model training for each use case,” he explains. “It worked for situations with uniform data and repeatable outcomes, like predictive maintenance or algorithmic trading. However, AECO is defined by complexity, fragmentation, and one-off projects. You cannot copy and paste insights from one site or building to the next.”
The advent of foundation models upends this logic. Trained on vast datasets and capable of generalised reasoning, these models can adapt to diverse contexts without being retrained from scratch. More significantly, they have become multimodal, capable of interpreting text, images, audio, video, and even sensor data in a single frame of understanding. This creates a new opportunity to bridge the messy real-world inputs of AECO with coherent digital outputs.
Geiger views this as a turning point. “We are no longer confined to narrow tasks,” he adds. “These models can read floorplans, reason about regulations, and suggest design alternatives. They do not just react to instructions; they engage with the task like a colleague would.”
The shift to agentic intelligence
AI is no longer just a tool that executes commands. The most recent models behave more like intelligent collaborators, capable of managing workflows, reasoning through complex tasks, and even self-correcting their steps. This agentic capability has profound implications for the design and construction lifecycle.
“Think of a future design workflow,” Geiger proposes. “You could task an agent to develop three building concepts based on client needs, local regulations, and energy performance criteria. It would generate floorplans, iterate based on performance simulations, and present outcomes for human review, all without manual coordination between siloed teams.”
This is more than hypothetical. Recent releases, such as OpenAI’s GPT-4 Turbo with tools and memory or Google’s Gemini 1.5, demonstrate the ability to manage long, multistep reasoning processes. Within Nemetschek, Geiger’s team is already testing these capabilities against AECO-specific tasks, from reading spatial layouts to parsing environmental requirements.
The company’s approach is to integrate this intelligence not as a separate platform but as an embedded layer throughout its portfolio of software brands. “We are building a group-wide AI layer that functions across applications,” he explains. “In some cases, it is visible as an AI assistant. In others, it works silently in the background to make things faster or more precise.”
That group-wide integration is no accident. Nemetschek’s AI and Data Innovation Hub, launched last year, is designed to accelerate innovation across the company’s diverse brands, from design and structural engineering to building operations and decommissioning. It is not just about embedding generative capabilities but embedding intelligence within processes, surfacing suggestions at the right time, and improving decision quality without increasing friction.
“The promise of AI is not just in what it can do, but how seamlessly it can do it,” Geiger explains. “If we get the workflow integration right, we can bring users with us rather than asking them to change how they work.”
Creativity augmented, not replaced
There is no suggestion that generative AI will eliminate the architect. If anything, the value of human creativity is being reaffirmed. Midjourney and other image models may generate dazzling images but lack the cultural, contextual, and tactile sensitivity that defines great design. “We believe architecture will always require human taste,” Geiger says. “AI can produce variations, optimise geometry, or analyse energy flows. But beauty, elegance, and cultural resonance are still human judgments. We aim to help professionals get to their best work faster.”
This philosophy is not just aspirational. It is grounded in recent research. A study by Harvard Business School found that individuals working with AI outperformed traditional teams in average quality of work. However, the highest share of exceptional results came from teams augmented by AI, suggesting that augmentation, not automation, is the true sweet spot.
For Nemetschek, this opens the door to democratising access to high-quality design. By reducing the learning curve associated with expert tools and capturing institutional knowledge in custom AI models, the company aims to lower the barrier to creative participation. “If an architectural firm can embed its design language in an AI model, onboarding new talent becomes much faster,” Geiger explains. “You are no longer waiting years for someone to absorb that culture. It is a form of knowledge transfer but at machine scale.”
This also creates a more inclusive future for the profession. By minimising the technical barriers to entry, AI has the potential to open the field to new voices and designers who may have been deterred by steep software learning curves or opaque planning processes.
Trust and transparency in deployment
If the creative promise of AI is real, so too are the ethical concerns. Architectural design is intellectual property. Clients demand confidentiality. Regulatory failure in construction is not just costly; it can also be deadly. This is why Geiger has placed trustworthy AI at the centre of Nemetschek’s strategy. “We run our models in a controlled environment,” he says. “Customer data is not passed to third-party model providers. We have a strict separation between generic company knowledge and project-specific content. The architect’s ideas are never used to train someone else’s model.”
This is more than data privacy. It is about control over outcomes. While AI is improving rapidly, it is still fallible. Geiger draws a parallel with computational design, another past frontier where outputs had to be verified rigorously. The same human-in-the-loop safeguards must apply here.
Transparency is key. Users need to know when AI is being used, what it generates, and who is responsible for validation. Geiger emphasises that Nemetschek’s tools will never obscure the role of AI. “We want people to trust the tool but question it when necessary,” he adds. The company has established internal principles focused on privacy, governance, and outcome accountability. These guidelines will underpin all new AI features, whether visible assistants or silent agents running behind the scenes.
The industry’s moment to move
The AECO sector cannot afford to wait. The capabilities now available through foundation models, such as reasoning, multimodal input, and embedded agents, are advancing too fast to ignore. Unlike previous technology cycles, the new wave of AI does not require perfect data or standardised processes to be effective. It can work with the ambiguity that defines this industry.
Geiger believes AECO is uniquely poised for transformation. “Everywhere I look, I see opportunity,” he concludes. “From generative design to simulation, from regulatory planning to construction scheduling. These models do not need to be trained on every building. They can reason, infer, and adapt.”
At Nemetschek, this vision is now operational. The AI and Data Innovation Hub acts as both an internal catalyst and external enabler, supporting software brands across the group while developing shared infrastructure and reusable agents. The aim is not to create one AI product but to ensure every product in the portfolio becomes AI-enabled.
The ambition is clear: Integrate intelligence where it matters most, empower professionals without displacing them, and bring the industry closer to a future where creativity and efficiency are no longer opposing forces but tightly integrated partners in the act of building.
