Can AI Improve the Future of Construction Takeoff?

[AlleyJoe]

Hello everyone, I am thinking a lot about future and I feel little worried because I don’t know if AI can really help in construction material takeoff or not. Many time when I do takeoff manual or even with normal software, mistake happen like missing items, wrong reading of drawing or not adding waste, and then my estimate go wrong. Some people say AI can read blueprints automatically and make more accurate result, but I am not sure if it is true or just hype. I really want to ask if you believe AI can replace human in construction material take off or it is only for support and still need manual checking. If anybody already try AI-based tool for takeoff please share your experience, because I feel serious tension that if I don’t improve my method now then my estimating will always stay weak and project will suffer.

 

[arwat23]

Some people say AI can read blueprints automatically and make more accurate result

Some people are delusional. I'm sure AI will get there eventually, but right now, it likely won't be all that accurate unless the data you feed it is perfect (and we all know it won't be perfect - it never is).

AI, especially current-gen AI, is a tool. It won't be able to completely replace people, at least not for a while. AI can help check things, make sure you're not missing something obvious or ask you questions to ensure your checking is correct, but it can't replace people for this specific task yet.

I'm sure it'll get there - AI is probably like the internet, it'll change the world forever - but right now there is just too much hype around AI. It's a bubble, and we won't know how useful it'll actually be until the bubble pops.

 

[arwat23]

I feel serious tension that if I don’t improve my method now then my estimating will always stay weak and project will suffer.

One way I use AI for my work is as a sanity check. I can give it info about a project and I'll ask it to spit out a list of questions I can use to double check my work, or I can give it a list of corrections and a floor plan and ask if I'm missing something. I wouldn't rely on it to do the work for me - it's no where close to being accurate and often tells me things that are flat out wrong, but I can use it to improve my own work, to make sure I don't overlook something obvious.

AI can definitely "read" a floor plan, but its comprehension of what it sees can be hit or miss (usually miss in my experience). If the plan isn't exactly what the AI is trained to look for, it can miss things. For example, I just had an AI review a accessible toilet room plan. Everything drawn was correct and compliant. The AI missed almost every clearance dimension and claimed that they needed to be added. It asked for additional requirements that either weren't applicable or not even what the jurisdiction uses.

In short, use AI as a tool to double check your work, not a tool to do your work. Don't blindly follow what it says, always double check what it tells you. No method of anything is perfect, every project is a lesson.

Try testing this out on a simple project with ChatGPT (preferably a paid plan since that gives you access to the usually better models). It isn't purpose built for this sort of stuff, but it can provide a decent example of current AI limitations.

 

[Yankee Chronicler]

Some people are delusional. I'm sure AI will get there eventually, but right now, it likely won't be all that accurate unless the data you feed it is perfect (and we all know it won't be perfect - it never is).

AI, especially current-gen AI, is a tool. It won't be able to completely replace people, at least not for a while. AI can help check things, make sure you're not missing something obvious or ask you questions to ensure your checking is correct, but it can't replace people for this specific task yet.

I'm sure it'll get there - AI is probably like the internet, it'll change the world forever - but right now there is just too much hype around AI. It's a bubble, and we won't know how useful it'll actually be until the bubble pops.

I agree. I think eventually AI will be able to do decently accurate take-offs -- IF the construction documents are decently accurate, which is a problem for another discussion.

I also think it will be many years before AI reaches a point where we can trust it.