Architects Sharpen Design Judgment for AI Era
In a June 17, 2026 Architizer essay, Nitsan Bartov, a PhD candidate at Henning Larsen and the Royal Danish Academy and an Ambassador for Architecture at Krea.ai, argues architects need an "architecture gym": deliberate manual reps such as sketching, model-making, and art history to keep design judgment sharp as generative AI enters practice. Bartov compares today's AI tools to a technically skilled but context-poor intern that can render a concrete building in timber with one prompt, producing a convincing image that may be structurally impossible. The piece's core warning, per the author, is not a sudden AI takeover but a slow, task-by-task erosion of design judgment if architects stop exercising the craft skills that let them catch a model's plausible-but-unbuildable output.
The useful signal here for AI tool builders, not just architects, is Bartov's framing of the risk: it is not sudden job displacement but a gradual, hard-to-notice transfer of judgment from person to model, task by task, until practitioners lose the ability to catch outputs that look right but are not buildable. That argues for AI tools in any expert domain to expose verifiable, checkable parameters rather than just polished output.
What happened
Nitsan Bartov, a PhD candidate at Henning Larsen and the Royal Danish Academy and an Ambassador for Architecture at Krea.ai, published an essay on Architizer proposing an "architecture gym": deliberate practice in sketching, model-making, art history, color theory, and composition to keep architects' design judgment sharp as generative AI tools spread through practice. Bartov describes current generative AI as akin to "a highly ambitious and capable intern with great technical skills and near-zero understanding," and gives an example of the risk: a building designed in concrete can be re-rendered in timber with a simple prompt, producing a visually convincing but potentially unbuildable result.
Industry context
Bartov's employer, Henning Larsen, announced an enterprise partnership with Krea in May 2026 to co-develop generative-image tools for architectural practice, choosing a bottom-up rollout that let architects experiment with AI tools before the firm committed. Bartov's essay reflects that same practitioner-first framing: the argument is that craft expertise (materials knowledge, construction experience, compositional training) functions as a verification layer against generative tools that optimize for visual plausibility over buildability.
For practitioners
Two implications follow for teams building AI tools for expert domains, architecture or otherwise: models and interfaces that expose controllable, checkable parameters (dimensions, structural constraints, material properties) will outperform tools optimized purely for visual appeal, and workflows that pair generative acceleration with deliberate maintenance of domain judgment will beat either pure automation or prompt engineering alone.
What to watch
- •Whether Krea and competing design-AI vendors add buildability checks, audit trails, or BIM-compatible export formats rather than image quality alone.
- •Training or continuing-education programs that pair AI-prompting practice with traditional craft exercises.
- •Further detail from the Henning Larsen-Krea enterprise partnership on how the firm measures judgment retention versus pure output speed.
Editorial analysis
This is a single-source practitioner essay from an author with a direct commercial tie to Krea.ai (Bartov is the platform's architecture ambassador), so the piece should be read as informed practitioner opinion advocating a specific tool-use philosophy rather than independent research; its central claims are not independently benchmarked.
Key Points
- 1Bartov proposes an 'architecture gym' of deliberate craft practice - sketching, model-making, art history - to keep design judgment sharp alongside AI tools.
- 2The described risk is gradual, task-by-task erosion of judgment, not sudden displacement, illustrated by AI rendering structurally impossible buildings convincingly.
- 3AI tools that expose checkable, buildability-relevant parameters will outperform tools optimized only for visually appealing output, per the essay.
Scoring Rationale
Solid-tier niche practitioner essay: genuinely useful design-judgment framing for anyone building AI tools in expert domains, but it is a single-source opinion piece by an author with a commercial tie to the AI vendor she discusses (Krea.ai Ambassador), with no new data or product news. Kept steady, flagged single-source with a commercial affiliation.
Sources
Public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
