40 Questions Architectsure and Built Environment Should Ask Before Trusting AI
When Autodesk AI generates five facade options or Rhino Grasshopper produces structural geometry, you have seconds to decide whether to proceed, refine, or reject. These questions help you catch the moments when AI has optimised for metrics that don't match what your building actually needs to do.
These are suggestions. Use the ones that fit your situation.
1When Midjourney or Autodesk AI shows you three design directions, can you identify which constraints it ignored because they were not in your prompt?
2If you deleted the AI-generated options and started sketching by hand, would your design move in a completely different direction?
3Does the AI output show you what is possible, or has it narrowed what you are willing to imagine?
4When you fed site photos and brief into generative design, did you check whether the AI understood the actual site conditions or invented a plausible version?
5Are you choosing between AI options, or are you using AI to explore options you already knew to test and modify?
6If a Revit AI suggestion for spatial layout conflicts with your design intent, can you explain why your choice is better without using the word 'preference'?
7Does the generated form solve a real design problem, or does it look interesting because the model was trained on award-winning buildings?
8Have you prototyped the AI-generated design at human scale, or only reviewed it on screen?
9When Grasshopper AI optimises a facade for solar gain, did it also account for privacy, maintenance, heritage context, and how it will age?
10Can you articulate what the AI got wrong before you show the output to your client?
Structural Judgement and Safety
11If you cannot hand-calculate the key structural moves in your building, can you confidently review an AI-generated structural model?
12When Revit AI or structural computation software produces calculations, does your team check the methodology or only the final numbers?
13Have you deliberately worked through a structural problem on paper recently, or has AI made that feel unnecessary?
14Does the AI output show you its assumptions about material properties, safety factors, and code compliance, or only the answer?
15If the AI has optimised a beam or column for weight, what did it sacrifice in terms of buildability, connection logic, or inspectability?
16When you spot an error in an AI-generated structural drawing, can you explain it to a structural engineer, or did you miss it because the drawing looked authoritative?
17Has your team named a person responsible for understanding and questioning every structural output from AI, or does responsibility scatter?
18If a contractor asks you why a structural member is sized the way it is, can you give a reason beyond 'the software calculated it'?
19Does your AI structural tool flag when it is making assumptions about loads, building use, or site conditions that might change?
20Have you tested whether your team would catch a 10 percent undersizing error in an AI-generated calculation, or does AI feel too authoritative to question?
BIM, Planning Analysis, and Optimisation
21When AI suggests changes to your BIM model to optimise cost or programme, does it understand your actual supply chains and contractor relationships?
22If Autodesk AI recommends a material substitution or system change, have you checked the source of that recommendation or only the outcome?
23Does your planning analysis tool show you which variables it weighted most heavily, or only the optimised result?
24When AI analyses your site plan for density, parking, or walkability, did it use local knowledge, or only algorithms trained on other cities?
25Has someone on your team manually checked whether the AI-optimised schedule is actually achievable with real suppliers and labour availability?
26If an AI tool says a design will cost X, can you trace back how it arrived at that figure, or is it a black box estimate?
27When Grasshopper generates multiple spatial or structural variants, are you comparing them properly, or just picking the one that looks best on screen?
28Does your BIM-linked AI tool account for the real cost and time of quality control, or only model-based metrics?
29If planning analysis shows that a site can support more density than you designed, have you challenged the assumptions, or accepted the AI output as ground truth?
30When you make a change to the BIM based on AI recommendation, do you understand the downstream effects on programme, cost, and constructability?
Human Purpose and Design Judgement
31Does your current use of AI tools prioritise speed and optimisation over understanding how people will actually experience the building?
32When you brief an AI tool, do you describe the social and human purpose of the space, or only the measurable constraints?
33If an AI-generated design scores high on efficiency metrics but feels inhuman, what would it take for you to reject it?
34Have you asked your team to deliberately design something the hard, slow way in the past month, or has AI replaced that practice?
35Does your practice still teach junior architects how to develop design intuition, or are they learning to brief AI and evaluate outputs?
36When a client chooses an AI-generated option over your sketched alternative, can you articulate what they would lose?
37If you removed all AI from your design process for one week, what would you rediscover about your own judgement?
38Does the building you are designing solve a problem for the community that uses it, or does it optimise for metrics the AI can measure?
39Have you tested how an AI-generated facade or spatial arrangement actually performs with real users and daylight, or only in simulation?
40When you present an AI-assisted design to a client, are you showing them your thinking process, or only the polished result?
How to use these questions
After using generative design or AI layout tools, sketch your own version without looking at the AI output. If it is fundamentally different, ask yourself why.
Assign one senior person to manually verify every structural calculation from AI. Rotate the responsibility so knowledge does not concentrate.
Before accepting an AI recommendation for cost, programme, or material substitution, ask the tool to show its working. If it cannot, treat the output as a starting point, not a conclusion.
Set a monthly minimum: one design decision made through slow, iterative thinking without AI. Use this to keep your own judgement sharp and catch what AI misses.
When you spot an error in AI output, document it and ask whether your team would have caught it. If the answer is no, retrain before you trust the tool again.