Areas measured by hand
Every estimate starts with a person measuring the areas from a PDF drawing by hand, using the dimensions and scale.
An anonymised case study of a timber-frame manufacturer. No names and no prices, only the market, its area-measuring problem and how it could work better.
This study is based on a real conversation with someone in the market. All data is anonymised.
Timber-frame manufacturers design and build timber-frame houses, roof trusses and cut timber in-house. Truss estimates are quick and there are many of them: they are calculated from the roof area, based on statistics of how much each roof shape costs per square metre. House estimates are detailed and fewer, with many nuances: different windows, roofing and facade cladding. Since there is no time to calculate everything, the more serious projects are pre-selected by the manager.
House projects arrive as PDF drawings. A person measures window, roof, exterior and interior wall and ceiling areas by hand from the dimensions and scale, then transfers the results into an Excel or Google Sheets estimate template. The estimating program also generates a material bill, the quantity of timber, the quantity and areas of metal plates, and that has to be typed line by line into the warehouse sheet so it is clear what there is and what needs to be bought.
General AI tools were tried for calculating areas from a drawing, but the results fluctuated and differed from the hand calculation. A language model on its own is not a reliable calculator, so an accurate result needs a different approach.
Every estimate starts with a person measuring the areas from a PDF drawing by hand, using the dimensions and scale.
Trusses alone come to about six hundred a year, plus a hundred-odd detailed house estimates.
The material bill for every project is re-typed line by line into the warehouse sheet.
General AI tools were tried for the areas, but their numbers fluctuated and did not match the hand calculation.
Estimates, clients and files live in different places and programs, so there is no single overview.
There are more enquiries than can be calculated, so some potential projects are left aside.
Imagine a tool where you upload a drawing and it marks the areas on its own and feeds them into verified formulas.
On a PDF drawing it marks walls, windows, doors, roof and ceilings in different colours and extracts the areas from the dimensions and scale.
Areas go into fixed formulas and are calculated by a real calculator, not by an AI guess. Where something is unclear, it flags which page to check.
From the material bill the quantities land in the warehouse on their own, with no line-by-line manual entry.
The result: more accurate areas, less manual work and a warehouse that updates itself instead of being typed in every time.
The system gathers the areas and the warehouse, and the person is left with the review and the decisions.
Formulas and a real calculator do the maths, so there are fewer errors than a general AI guessing.
The warehouse fills from the material bill, so it is always clear what there is and what to buy.
Around six hundred truss estimates and a hundred to a hundred and twenty house estimates are produced a year, and every area is measured from a PDF by hand.
General AI tools fail here because the scale and the notation of the drawings are specific. The only thing that works is a solution where AI recognises and marks up, while the company's own formulas do the calculating.
Areas are recognised and marked up automatically, formulas calculate the quantities, and the stock is filled from the same bill of quantities.
Less measuring by hand, more precise areas, and a stock record that matches what was actually calculated.
You can send this summary to a colleague or save it as a PDF.
If every estimate also starts with measuring areas by hand for you, let's talk. Together we will look at how this could work in your company.