All MicroEvals
Act as an independent, evidence-led research analyst special...
Create MicroEval

Act as an independent, evidence-led research analyst special...

Prompt

Act as an independent, evidence-led research analyst specializing in architecture, AEC technology, AI adoption, and agentic systems. RESEARCH TASK Identify architecture firms in the United States and internationally that have publicly documented the use, pilot, or development of AI in their architectural workflows. Then provide case studies relevant to AI or agentic AI in architecture services. This is a research task, not a marketing overview. Distinguish evidence of actual use from research, announcements, and vendor claims. Be candid if public evidence for agentic AI in architecture is limited. SCOPE Focus on architecture practices that provide architectural design services. Multidisciplinary architecture and engineering firms may be included, but label them clearly. Do not count architecture software vendors, AI startups, developers, contractors, or engineering-only consultancies as architecture practices. If useful, list vendors separately in a clearly labeled section; do not present them as architecture firms. Aim to identify 10–15 architecture practices, including firms based in the United States and firms outside the United States across multiple world regions. Include fewer if the available evidence does not support a larger list. Do not invent examples to meet a target. DEFINITIONS AND CLASSIFICATION For every firm or case, distinguish the following: AI adoption status Use one of these labels: - Documented operational use: Public evidence describes AI used in a live company workflow or project. - Pilot or limited deployment: Public evidence describes a test or limited use, not established routine deployment. - Research or prototype: Research, experimentation, or a demonstration, without evidence of production use. - Announced or unsubstantiated claim: A claim is public, but the operational details or evidence are unclear. - Not verified: You could not establish a reliable public source for the claim. Do not treat a firm’s general interest in AI, AI hiring, or AI-related commentary as proof of workflow integration. Agentic AI Use the term agentic AI narrowly. A case qualifies as agentic only when public evidence supports a goal-directed process that performs multiple steps, interacts with software or tools, and uses some form of feedback, validation, or iteration. Identify the evidence for each feature. Use these maturity labels: - A0 — Conventional automation: Rules, scripts, macros, or workflows without AI. - A1 — Single-task AI assistance: A bounded AI task, such as generating text, images, analysis, or a suggestion. - A2 — Multi-step but human-directed AI workflow: AI helps with several steps, but the user directs the sequence and decisions. - A3 — Bounded agent: The system can plan or sequence actions, use tools, and respond to results, with defined human oversight. - A4 — Multi-stage agentic workflow: An agent coordinates multiple workflow stages with limited human intervention and clear controls. Only classify a case as agentic if evidence supports A3 or A4. Label A0–A2 examples as AI or automation, not as confirmed agentic AI. A chatbot, generative image tool, parametric design system, optimizer, or one-shot layout generator is not automatically an agent. RESEARCH AND EVIDENCE RULES 1. Use web or research tools if available. State the research “as of” date. 2. If you lack live browsing or cannot verify sources, say so prominently. Do not imply that you performed live research. 3. Prefer project-specific and workflow-specific evidence, such as: - Architecture-firm case studies or technical publications - Named client or project documentation - Peer-reviewed papers, conference papers, or research reports - Reliable industry publications 4. Use vendor material when relevant, but label it as vendor-published or vendor-supplied evidence. 5. Provide citations close to the claims they support. For each source, give its title, publisher, publication date if available, and direct URL. 6. Do not cite a homepage as evidence for a specific project or AI capability. 7. Do not invent company deployments, projects, sources, dates, tools, performance figures, or quotations. 8. Only report quantified results when a cited source provides them. Explain the reported measure and context; do not imply that a vendor claim is independently validated. 9. For significant case studies, seek both a primary source and independent corroboration where available. State when only a company or vendor source is available. 10. If evidence does not support any publicly documented A3/A4 architecture case, say so. You may describe the closest AI-assisted examples, but clearly label them as non-agentic or unverified. REQUIRED RESPONSE FORMAT 1. Research date and limitations State the research date, whether you had live web access, and any important evidence limitations. 2. Executive summary Summarize the landscape and state how much publicly documented evidence you found for genuinely agentic AI in architecture services. Do not overstate the maturity of the field. 3. Method and definitions Briefly explain how you distinguished operational use, pilots, research, and marketing claims—and how you classified agentic AI. 4. Architecture-firm landscape Provide a table with these columns: | Firm | Headquarters / region | Architecture-practice type | Specific AI workflow or project | Adoption status | AI category | Agentic level | Evidence and sources | Include a concise, specific use case for each firm. Do not list a firm solely because it has discussed AI generally. Aim for at least four US-based firms and six firms outside the United States across at least three other regions, where source quality allows. If the evidence does not support this, provide fewer and explain why. 5. Case studies Provide 3–6 case studies if reliable evidence exists. Include at least one genuinely agentic case only if evidence supports it. If no confirmed public A3/A4 case is found, state that clearly and describe the closest relevant examples as non-agentic or unverified. For each case study, cover: - Firm, project or workflow, location, and date - The architectural problem or task - Inputs, AI or software actions, and outputs - What the AI actually did versus what a human did - Whether it used tools, performed multiple steps, and checked or responded to results - Adoption status and agentic level, with reasons - Reported outcomes, if any, and how they were measured - Evidence quality, source limitations, and what is not known - Direct citations Do not call a project “agentic” solely because a source uses that word. 6. Agentic-AI evidence summary Provide a short table of the strongest agentic claims and classify each as: - Publicly supported A3/A4 - A2 or agentic-adjacent - AI-assisted but non-agentic - Not sufficiently evidenced For each, identify which agentic capabilities are supported by evidence and which remain unverified. 7. Practical implications Give 3–5 concise implications for an architecture firm considering AI adoption. If discussing a CAD-to-3D-to-render workflow such as AutoCAD, SketchUp, Revit and Lumion, label the points as recommendations—not as evidence about the firms listed. 8. Sources List the sources used, with title, publisher, date if available, and direct URL. Identify sources that are company- or vendor-published. STYLE Be clear, concise, and specific. Prefer a smaller, well-supported list to a larger speculative one. Separate verified facts, claims made by companies or vendors, and your own interpretation.

Drag to resize

Response not available

Drag to resize
Drag to resize