When I consulted AI about 'the estimate being too high,' it exposed the darkness of 'multi-layered subcontracting' in the construction industry.
"I want to build a good building. But the budget is limited."
This is a worry shared by every client.
Especially in the last few years, the surge in construction material costs has been abnormal.
It is not uncommon for a project planned a few years ago to balloon to 1.5 times the cost when you finally get an estimate.
At such times, we architects and site supervisors perform cost reduction adjustments called 'VE (Value Engineering).'
Changing wall materials to cheaper ones, replacing floor tiles with sheets, reducing the number of shelves...
We swallow our tears, lower the 'quality,' and try to somehow keep it within budget.
But is the 'quality of the building' really what should be cut?
What if there is 'wasted money that doesn't benefit the client' lurking inside the estimate?
This time, I would like to talk about how letting Google's generative AI, 'NotebookLM,' read an estimate exposed the bad practices of the construction industry—the 'waste of multi-layered subcontracting'—to the light of day.
AI walks through a 'maze of numbers' with thousands of lines
This happened during a new office building project.
As expected, the estimate from the construction company was significantly over budget.
Normally, this is when you start worrying about 'where to make it cheaper (where to lower specifications).'
But this time, I started by doubting, 'Is this estimated unit price even appropriate in the first place?'
I had an estimate as thick as a phone book in my hand.
It would take days for a human to check the unit price of every single item.
I fed the PDF into the AI (NotebookLM) and gave it this instruction:
"Compare the unit prices in this estimate with standard market design unit prices and list the items with large discrepancies."
The AI, without a single complaint, instantly read through thousands of lines of data and, a few seconds later, spat out 'a certain sense of discomfort.'
The culprit was inside the 'large-format tiles'
The AI's answer was as follows:
> "Warning: The construction cost for 'large-format tiles' in the exterior wall work is calculated at approximately 1.5 to 1.8 times the general market price. Material costs are at list price, but labor costs are abnormally inflated."
>
This tile work alone accounts for a large weight of the total construction cost. If this becomes a fair price, it will be a significant reduction.
Why is it so expensive?
Is it a simple rip-off? Or is the construction extremely difficult?
Armed with the AI's findings, we conducted a hearing with the construction company.
"Isn't this unit price too high compared to the market rate? Please tell us the basis for it."
What was revealed there was the structure of the 'telephone game' that remains deeply rooted in the construction industry.
Two 'ghost companies' sandwiched in between
As we dug deeper, a surprising commercial flow (the flow of goods and money) emerged.
Originally, that tile manufacturer could deal directly with the construction company.
However, the construction company this time had no 'experience handling those tiles.'
* The construction company requested an estimate from 'Subcontractor A,' with whom they had a relationship, just in case.
* But since Company A also had no experience handling them, they passed it to 'Subcontractor B.'
* Company B was also not confident, so they passed it to 'Subcontractor C' (the company that actually does the construction).
* Company C got an estimate for materials and labor from the manufacturer.
As a result, there were two companies sandwiched between the manufacturer and the construction company that were 'just passing the estimate from right to left.'
Moreover, they didn't make it expensive out of malice.
"It's a special tile we've never handled before. What if it breaks? What if the delivery is late? Just in case, let's add a little extra profit (risk hedge)."
As a result of each company adding on 'fear fees' (safety margins) like that, the final estimate had ballooned to 1.8 times the market rate.
No one is doing the construction, yet only the money keeps piling up. This is the true nature of 'multi-layered subcontracting.'
By eliminating the 'middleman,' we were able to reduce costs without lowering quality
Once the cause is known, the solution is simple.
We demanded this correction from the construction company.
"Please do not go through the two companies in between, and place the order directly with the manufacturer's responsible construction."
We organized the commercial flow and eliminated the telephone game.
As a result, we succeeded in reducing costs to a fair price without changing the grade (appearance) of the tiles at all.
If the AI hadn't found it, we might have made the choice to lower the quality of the building by saying,
"There is no budget, so let's change to cheaper siding."
That is an unfortunate outcome for both the client and the designer.
AI also doesn't miss 'miscounts' in quantitiesBy the way, the AI found not only unit price errors but also 'quantity' mistakes.
> "Compared to the wall area of the conference room calculated from the drawings, the wallpaper quantity in the estimate is calculated about 20% higher."
>
The AI coldly points out errors that might have been missed if a human had been checking with a calculator.
We corrected this as well and reduced unnecessary costs.
Architects should carry a 'protective sword'Cost reduction that just involves bowing your head and saying, 'Make it cheaper,' is not professional work.
That could lead to bullying subcontractors and also increases the risk of quality accidents (shoddy construction).
However, 'unfounded markups' and 'structural waste' should be thoroughly eliminated.
Until now, architects have been busy with design and drawing creation, and sometimes haven't been able to devote enough time to such 'scrutiny of estimates.'
But now, we have the strongest partner: AI.
"The estimate is too high and I can't accept it."
"I don't know where to cut."
What should be cut is not the 'quality' of the building, but the 'excess fat' in hidden places.
▼ For those who want to know more"How specifically do you let NotebookLM read an estimate?"
"I want a template for prompts (instruction text)."
Such practical know-how is explained on our office's official blog.
Whether you are in the same industry or are thinking about building in the future, please take a look.
📖Read the article: [ConstructionDX] Is that construction cost estimate too high? Spot hidden costs withAI (KTXArchilab OfficialHP )


[Author of this article Tetsuya Matsumoto]
Representative of KTX archiLAB, First-Class Architect, Part-time Lecturer at Osaka University of Arts
Ranked 9th in the Italian DAC Certified World Designer Rankings (Architecture Category) (1st in Japan)
▶Wikipedia Tetsuya Matsumoto (Architect)
[Contact Information]
KTX archiLAB First-Class Architect Office
S-Scenario Minami-Azabu 002, 3-4-5 Minami-Azabu, Minato-ku, Tokyo
Nisshin Building 2F, 298-2 Funaoka-cho, Himeji-shi, Hyogo
(Business Description)
Architectural Design / Interior Design
Email: kentixx@ktx.space
Phone Number: 03-4400-4529 (Main)
Website:https://ktx.space/
[Related Article Links]
▶Representative Matsumoto Newsweek Magazine Interview
▶KTX archiLAB Profile (Company Overview)
▶Architectural Design & Interior Design Case Studies
▶KTX archiLAB YouTube Video Channel
いいなと思ったら応援しよう!
この記事は noteマネー にピックアップされました

