Henning Larsen Partners with Krea to Embed Imagery

Design practices embedding generative-image tools into routine workflows typically move imagery earlier in the design lifecycle, converting renders into working communication artifacts used for alignment, critique, and rapid iteration. That shift matters to practitioners because it changes pipeline constraints: verification, asset provenance, prompt/version control, and BIM handoffs become operational concerns rather than optional niceties.

Per Krea's company blog, Krea announced an enterprise partnership with Henning Larsen on May 29, 2026 (Diego Rodriguez, Krea blog). Architizer's coverage describes Henning Larsen's adoption as deliberately bottom-up: staff across geographies and roles tested multiple AI tools in a sandbox before the studio engaged Krea, and Architizer quotes Eliana Nigro, Head of Digital Adoption, framing the process as "collective learning." Krea's blog notes that an internal survey shown to Diego Rodriguez indicated studio members had already chosen Krea as their preferred AI creative tool. Architizer reports practitioners using Krea in image-to-image workflows that take geometry exported from Rhino and Revit and translate it into imagery that carries atmosphere and intent. Architizer also records a practitioner quote: "AI can produce a beautiful image of a building that could never actually stand up," attributed to a Henning Larsen colleague.

Architizer documents a concrete workflow: teams export trustworthy geometry from CAD/BIM tools, then use Krea to render experiential visuals that communicate lighting, materials, and atmosphere beyond what technical drawings convey. Editorial analysis: across design firms, this model -> image pattern reduces ambiguity during early reviews but increases the need for verification layers that reconcile visual suggestions with structural, code, and engineering constraints.

Industry-pattern observations: firms that democratize access to generative tools face three operational pressure points. First, asset and prompt provenance, logging model versions, seed/prompt text, and source geometry, becomes essential for reproducibility and audit. Second, cross-disciplinary review cycles must include explicit checks where visual outputs imply structural conditions; Architizer's quoted warning about nonbuildable imagery shows this risk in practice. Third, integration with BIM/CAD workflows matters: maintaining a clear mapping between a final image and the originating Rhino/Revit model helps downstream engineering and costing.

Observers should track:

• •whether vendor partnerships extend to native BIM integrations or export-safe formats
• •whether studios publish prompt-and-model registries or style-libraries for reproducibility
• •how firms govern visual outputs that suggest infeasible designs. These indicators will show whether studio adoption stays at exploratory communication use, or requires formal QA and compliance processes

Reporting by Krea and Architizer shows a practical, design-focused adoption pattern: a bottom-up rollout, image-to-image use of trusted geometry, and a cultural emphasis on shared literacy rather than siloed specialist control. Editorial analysis: for ML engineers and data teams working with creative studios, priorities shift from raw model capability to integration, provenance tooling, and human-in-the-loop verification.