From “AI sandbox” to default workflow: how Gensler is operationalizing generative design at scale
Gensler’s three-year experiment with generative AI—incubated in an internal “AI sandbox” and now deployed across the majority of its roughly 3,000 projects per year—signals a pivotal shift in how large architecture practices are turning AI from novelty into infrastructure. Under Co-CEO Jordan Goldstein, the firm is blending third‑party models with proprietary tooling to accelerate early-stage design and sharpen decision-making around performance, experience, and feasibility.
What stands out is not simply faster rendering or more dramatic visuals, but the expansion of AI into the analytical core of architectural work. By simulating natural light, acoustics, occupancy patterns, and energy performance earlier in the process, AI becomes a mechanism for compressing the time between an idea and a defensible design direction. That compression matters commercially: early phases often carry high uncertainty, heavy iteration, and significant labor costs—precisely where automation and computational exploration can have outsized impact.
The firm’s reported ability to evaluate roughly three times as many design permutations per project underscores a broader industry reality: generative design is evolving into a force multiplier. It does not replace the architectural brief or the human-led synthesis of constraints; it expands the search space, making it more likely that teams identify high-performing options before budgets, schedules, and stakeholder politics harden into immovable constraints.
Performance-first AI in the built environment: what the flagship projects reveal
Two cited applications illustrate how AI is being positioned less as an aesthetic engine and more as a systems intelligence layer for the built environment.
- Under Armour’s Baltimore headquarters: AI was used to model employee movement and infer implications for ventilation and airflow needs. This is an important tell. Occupant behavior—how people actually move, cluster, and use space—has historically been approximated through rules of thumb, post-occupancy studies, or expensive bespoke modeling. AI-enabled occupancy simulation brings that analysis forward, potentially improving comfort, health outcomes, and operational efficiency while reducing redesign risk later.
- Baghdad Sustainable Forests (Iraq): Positioned as the Middle East’s largest ecological development, the project highlights AI’s alignment with ESG and decarbonization mandates. Large-scale ecological and climate-adaptive developments are inherently multi-variable: microclimates, water, shading, materials, mobility, and long-term maintenance all interact. AI’s value proposition here is not just optimization, but the ability to test scenarios and communicate trade-offs credibly to governments, investors, and communities.
Equally consequential is the use of AI for cinematic narratives that improve client buy-in. Architecture has always required translation—turning technical intent into stakeholder confidence. As generative AI converges with VR/AR, digital twins, and real-time simulation, the industry is moving toward end-to-end experiences where clients can “feel” operational outcomes (light, sound, circulation) rather than interpret static drawings. In a market where approvals and financing often hinge on clarity and persuasion, storytelling becomes a strategic capability, not a cosmetic add-on.
Competitive economics, talent redesign, and the rise of proprietary AI stacks
Gensler’s approach points to a strategic playbook other large firms are likely to emulate: build a modular, proprietary interface that can orchestrate multiple generative engines rather than depend on a single vendor’s monolithic platform. This matters for three reasons:
- Data control and confidentiality: Architecture projects routinely involve sensitive tenant data, security considerations, and commercially confidential plans. A firm-controlled stack can better enforce access controls, logging, and compartmentalization.
- IP governance and customization: Proprietary layers can encode firm-specific standards, design heuristics, and reusable components—turning institutional knowledge into scalable software advantage.
- Client lock-in across the asset lifecycle: Integrated platforms that connect concept design to BIM, simulation, and facilities management create continuity that can extend relationships beyond handover.
Economically, AI can reduce labor intensity in preliminary phases—historically a major share of billable effort. That creates strategic options: firms can price more aggressively, protect margins, or reinvest into higher-value advisory services such as sustainability strategy, workplace analytics, or performance verification. Over time, this could reshape the competitive landscape, rewarding practices that treat AI not as a tool purchase but as an operating model.
The talent implications are equally structural. As routine drafting and baseline calculations become increasingly automated, junior roles may shift toward:
- AI supervision and validation
- computational design and data literacy
- prompting and workflow orchestration
- ethics, security, and model-risk awareness
This is less about turning architects into software engineers and more about making AI fluency “table stakes” for creative leadership—so that human judgment remains the differentiator rather than an afterthought.
Trust, authorship, and governance: the industry’s unresolved fault line
The most important counterweight to AI enthusiasm is trust. A 2025 American Institute of Architects study reporting that 90% of architects have concerns about accuracy, security, and intellectual-authorship risks captures the profession’s central tension: AI can accelerate exploration, but it can also blur provenance, amplify errors, and weaken accountability if adopted without guardrails.
Academic perspectives reinforce this duality. Pratt Institute’s Jason Vigneri-Beane highlights AI’s strength in rapid prototyping while warning about erosion of creative agency; UNC Charlotte’s Sabri Gokmen argues AI is likely to underpin 3D geometric and data-driven design, with humans retaining the decisive role in judgment. Taken together, these views suggest the next phase of adoption will be less about capability and more about governance—how firms validate outputs, document decisions, and attribute authorship in ways that withstand legal scrutiny and professional ethics.
For architecture and construction—an industry tied to safety, regulation, and long-lived assets—AI’s future will be determined by who can operationalize trust at scale. The firms that win are likely to be those that pair generative speed with rigorous model validation, secure data environments, and transparent attribution practices, turning AI from a creative accelerator into a dependable engine of performance, accountability, and long-term value.
By
By
By
By
By
By
