AI in construction: from site plan to bill of quantities

Before a single block is laid, a building project lives or dies on its estimate. The bill of quantities, a line-by-line list of every material and task with its quantity and price, decides what the project costs, how it is financed and how suppliers are paid. Preparing one by hand takes days. Getting it wrong can sink a project.

This article looks at why construction has been so slow to change, where AI is starting to help, and what a responsible tool should and should not do.

An industry that has barely changed

McKinsey Global Institute's landmark study of construction productivity found that labour productivity in the sector grew by only about 1% a year over the two decades before 2017, compared with 2.8% for the whole world economy and 3.6% for manufacturing. Construction accounts for around US$10 trillion of spending a year and employs about 7% of the world's working-age population, so that gap matters.

The same research estimated that acting on a set of improvements together, including better design, procurement and digital tools, could raise productivity by 50% to 60%.

Projects run late and over budget

Overruns are the norm, not the exception. Bent Flyvbjerg's database of more than 16,000 large projects, described in the book How Big Things Get Done, found that over 90% went over budget, over schedule or both, and fewer than 1% delivered on time, on budget and with the promised benefits.

The consultancy HKA's analysis of 1,800 disputed projects in 106 countries found that incorrect, incomplete or late design was a factor in nearly 45% of them, and that schedule extensions were especially severe in the Middle East and Africa. Many of these problems begin with poor information at the start.

Africa needs to build fast

The scale of need is enormous. UN-Habitat figures reported in 2025 put Africa's housing deficit at around 70 million homes, and more than half of urban residents in Sub-Saharan Africa live in informal settlements. Meeting that need means building more, building faster and building within budget, often for families who are financing a home stage by stage or sending money from abroad.

Where AI helps today

Quantity take-off

Measuring quantities from drawings, how many square metres of wall, how many cubic metres of concrete, is repetitive and error-prone. Software that reads plans and models can do much of it automatically. One vendor-hosted study of an AI take-off tool reported about 70% time savings with results within 5% of a manual take-off, though it involved a single user and should be read as an early signal rather than proof. Academic work on automated quantity take-off from building models has also found large time savings and caught significant errors in manual figures.

Turning a plan into a priced estimate

Once quantities are known, standard construction rules can turn them into materials and labour, and a price list turns them into money. Doing this in minutes instead of days means a builder can give a client an estimate the same day, and update it when prices change.

Keeping prices current

Material prices can change quickly, especially where currencies move. Connecting estimates to suppliers who publish their prices keeps a bill of quantities honest.

Tracking progress

Photos, locations and inspection records for each stage let clients, lenders and suppliers see what has actually been built. That makes it easier to release money stage by stage, and harder for problems to hide.

What AI should not do alone

• Structural decisions. Foundations, beams and loads must be designed and checked by qualified engineers. AI can assist, flag and calculate, but a professional must sign off.
• Final quantities for contracts. A registered quantity surveyor should check the rules and outputs, especially when money changes hands on the basis of the bill.
• Local regulations. Building codes and approvals differ by country and city. Tools must be checked against local rules.
• Guesswork on prices. An estimate is only as good as its price list. Placeholder prices must be clearly marked and replaced with real supplier prices.

The independent evidence on AI estimating accuracy is still thin. Treat any tool's accuracy claims as something to test on your own past projects before relying on them.

How we are approaching it

This is the problem Mabwezi is being built to solve. It starts with fast, priced bills of quantities for common building types, connects builders with suppliers who publish their prices, supports staged finance once each stage passes inspection, and gives clients, including families abroad, a private view of their build. The roadmap extends from a site plan to drawings, structural checks and renders, always with qualified engineers and quantity surveyors in the loop.

Construction will not be transformed by a single clever model. It will change the way most industries have changed: by taking the slow, repetitive parts of the work, doing them faster and more consistently, and giving skilled people better information to make the decisions that matter.