Gensler Deploys AI Across Design Workflows
Business Insider reports that architecture firm Gensler created an "AI sandbox" three years ago to test generative tools, according to co-CEO Jordan Goldstein. Goldstein is quoted saying the firm could not be "reactive to artificial intelligence" and that AI "helps bring more ideas into the process" (Business Insider). The outlet reports Gensler works on about 3,000 projects a year, and that AI is used in some capacity on the majority of those projects. Business Insider describes the sandbox tests as involving vendor-created generative AI tools, some of which have been integrated into an in-house interface. The article gives an example where designers used AI to simulate occupancy and environmental performance for Under Armour's new headquarters in Baltimore, and notes AI can model sunlight, sound, and people flow during early concept work (Business Insider).
What happened
Business Insider reports that architecture firm Gensler began an internal initiative it calls an "AI sandbox" three years ago to experiment with generative artificial intelligence, citing comments from co-CEO Jordan Goldstein. Business Insider quotes Goldstein: "It was a key moment for us as a firm," and "What we've found is that AI really helps bring more ideas into the process and enables teams to explore their ideas more effectively." The article states Gensler handles about 3,000 projects a year and uses AI in some capacity on the majority of those projects. Business Insider describes the sandbox as a period of vendor-tool testing and reports that several vendor-created generative tools were integrated into an in-house interface. The piece gives a project example where designers used AI to simulate occupancy-driven environmental performance for Under Armour's new headquarters in Baltimore (Business Insider).
Editorial analysis - technical context
Generative AI in architecture is being applied to early-stage concepting tasks that are traditionally iterative and visual. Industry-pattern observations show these tools are commonly used to accelerate exploration of configurations such as sunlight studies, acoustic behavior, and circulation patterns, as Business Insider documents in Gensler's use cases. For ML practitioners integrating similar capabilities, the primary technical surface is usually rapid scenario generation plus simulation inputs, not turnkey building-control systems. Models are therefore paired with domain simulations or rule-based checks to translate generative outputs into testable design alternatives.
Industry context
Firms with broad project portfolios can scale learning from experiments into standardised toolchains, which Business Insider frames as the path Gensler followed by moving vendor models into an internal interface. Industry observers note that architecture and engineering practices increasingly treat generative models as components in hybrid pipelines where physics-based simulation and code-driven analysis verify generative proposals before they reach clients or builders.
For practitioners - what to watch
Observers should track three indicators: uptake of vendor models into firm-level UIs; coupling of generative outputs with validated simulation engines; and documentation or templates that convert AI-generated concepts into downstream deliverables. Business Insider does not quote Gensler on long-term governance or vendor relationships, and Business Insider does not report that the firm has published a technical whitepaper on the sandbox. For teams evaluating similar deployments, the article provides concrete examples of use cases and suggests practical entry points for prototyping generative-simulation workflows.
Scoring Rationale
The story shows a large, influential architecture firm operationalising generative AI across thousands of projects, a notable institutional case study for practitioners. It is relevant to teams building design-focused ML pipelines but does not introduce new models or generalisable open-source tools.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems


