A panel of architects and technology consultants gathered for an Architects’ Journal webinar to examine how AI for architectural practice is moving beyond generative imagery and into the operational fabric of offices, from knowledge retrieval and building code comparison to feasibility modelling and post-occupancy data analysis.
The webinar, chaired by AJ competitions editor Merlin Fulcher and supported by CMap, brought together Martha Tsigkari, senior partner and head of the Applied R&D group at Foster + Partners; Rada Daleva, architectural designer and founder of Daleva Design; Kira Ariskina, architect, founder and chief executive of ViableSite; and Phil Sanders, senior consultant at CMap.
AI for architectural practice at Foster + Partners: years of groundwork
Foster + Partners, the global studio for sustainable architecture, urbanism, engineering and design founded by Norman Foster in 1967, has been developing AI applications internally since 2018. Tsigkari was direct about the scope of that work. ‘We have been doing AI at Foster + Partners for many, many years. We started in 2018,’ she said.
The practice has built an application called Ask Foster + Partners, which allows staff to search across six decades of accumulated project knowledge, including documentation, images, materials and two-dimensional drawings. ‘We started by sharing our 60 years’ worth of knowledge through applications like Ask Foster + Partners,’ Tsigkari explained. The ambition extends further: different datasets are being connected and made accessible through AI agents, so that architects can ask questions and surface relevant information across the practice’s full project history.
Applied R&D at the practice also covers building code comparison, document translation, file search, report writing and project management tools that collect queries from across a project and route them to the appropriate people. AI is also being applied to operational data from completed buildings, with predictive models used for resourcing, cost profiles and profitability. The underlying applications are typically built on existing models and fine-tuned with the practice’s own data, rather than developed from scratch.
On design, Tsigkari described experiments in AI-generated layouts, parametric modelling and design optioneering, with designers able to bring sketches, images and three-dimensional models into the practice’s AI portal. When Fulcher raised the question of whether this made the practice’s offer less distinctive, her answer focused on culture rather than tooling. ‘The identity of Foster + Partners is not driven by any single tool, it’s driven by the culture at Foster + Partners and the design excellence that we are working towards,’ she said. Governance follows the same logic: AI-generated outputs go through the same checking process as a Revit model. ‘Technology is just a facilitator,’ she said. ‘It’s a means to an end, it’s not the end itself.’
Smaller practices, sharper questions
Rada Daleva, founder of Daleva Design, offered a perspective from a smaller studio operating under tight client programmes. Her starting point was a telling phrase from a recent client: ‘Fix this with AI, please.’ Daleva argued that this reflects a widespread misunderstanding of what the technology does. The design decisions, she said, remain with the architect. ‘I have fixed it, but through design,’ she said, describing choices about materials, depth, furniture and spatial relationships that AI helped her communicate faster, not make. ‘AI removed the cost of making a lot of options, but not the cost of choosing the right options,’ she said.
Her conclusion was that AI makes it possible for smaller teams to undertake more ambitious work, but raises the stakes on professional judgement rather than lowering them. ‘The question is not, is it faster, but is it better?’ she said. Current image-to-three-dimensional tools, she noted, do not yet provide the level of detail required for manufacture, so production elements were modelled in Rhino.
Kira Ariskina, founder and chief executive of ViableSite, brought a planning and data perspective. She is developing an AI-enabled platform for early-stage feasibility work on small urban sites, and her argument was that generic large language models lack the domain-specific knowledge that makes architectural judgement valuable. As routine drafting and summarising tasks become more widely automated, she suggested, the scarce resource becomes expertise. ‘The question kind of moves from, can I use AI, to what do I know that LLM doesn’t know?’ she said.
Ariskina challenged practices to record not only final documents but the reasoning behind decisions, so that historic project data becomes genuinely useful rather than merely stored. She also flagged the risks directly: ‘Please remember, rubbish in, rubbish out.’ Poor-quality data processed by an AI system can produce authoritative-sounding but fundamentally wrong answers. Liability, she stressed, remains with the qualified professional. ‘At the end of the day, you will be liable for those mistakes,’ she said. She also advised practices to check the terms and conditions of any large language model they use, particularly regarding how data shared internally may be passed to third parties.
Phil Sanders, senior consultant at CMap, addressed the practice management dimension. CMap is developing AI agents that can surface previous project performance when preparing fee proposals, flag client payment behaviour at invoicing, and update resourcing when a project stalls. Sanders described CMap Chat, a conversational interface allowing users to interrogate practice data in plain English. He also outlined a Model Context Protocol as a mechanism for connecting information held across separate systems, potentially allowing brand guidelines, contract terms and project records to be drawn together when preparing a proposal. ‘AI, for us, isn’t about replacing expertise, it’s about making it easier to use what we already know,’ he said.
The closing discussion turned to fees and professional value. All four panellists questioned whether billing by the hour remains the right model when AI compresses the time taken for certain tasks. Tsigkari argued the profession needs to articulate its value in terms of the final asset, the quality of the building and the quality of the drawings, rather than headcount and hours. Daleva suggested that competing on lower fees has already weakened the profession, and that AI should prompt practices to rethink how they structure and present their services. Ariskina was direct about construction costs: ‘Building costs are still happening in real life,’ she said. AI can accelerate communication within the design team, but it does not, in her view, alter the cost of building.
