Where is AI working in the lumber industry, and where isn't it?

AI is working in lumber on the cut: grading boards, detecting defects, deciding where to saw. In a primary mill that is a sensible place for it, because the cut decides what the log becomes. Where it is barely at work is in the scheduling around the cut in secondary manufacturing: sequencing orders to cut setups, changeovers and knife changes. The research on that problem is large and old. What lags is practice, and in remanufacturing those costs can be larger than the cutting.

What the usual answer says

The top results for this question describe the cut. Machine vision grades boards. Deep learning finds knots, cracks and wane. Optimisers choose the best cut from each log or board. Predictive maintenance watches the machines. Planning and sales are usually mentioned as the next frontier. That is accurate about where the products are, and it leaves out where costs sit in operations that are not primary mills.

Where the evidence is strong

A team led by Min Ji published an automated structural timber grading line in Scientific Reports in December 2025. They set out the problem it solves: "The anisotropic nature of wood renders standard visual grading a challenging task, often resulting in a grading process with low repeatability". Their line combined machine vision, moisture measurement and mechanical stress grading. The strongest evidence it works is that it was certified under a national grading standard in 2023.

A 2024 study in Sensors of a two-sided defect detection and cutting system reported "a commendable mean average feature detection precision of 0.94 when evaluated on a meticulously curated dataset comprising 450 images". The authors report a 12.3% gain in volume yield, on 100 pieces, from 69.2% with conventional fixed-length cutting to 81.5%. That is 12.3 points against the simplest cutting method, in one setting, not against an optimised line.

Where it isn't working yet

Scheduling with setups is not a neglected research problem. Kuo-Ching Ying and colleagues, in a 2025 review, considered "over 2100 articles published between 1986 and 2024" on scheduling with sequence-dependent setup times. Practice is another matter. A 2024 study of sawmill order scheduling by Francisco Vergara and colleagues, in Maderas, observed that "in most cases static heuristics, such as earliest due date (E), longest processing time (L), and shortest processing time (S), are used because of their simplicity". Rules like those don't take setups into account.

In secondary manufacturing the setups can be the larger cost. From Quarri's own work with a lumber and millwork manufacturer: setup, changeover and knife changes made up about 45%+ of one remanufacturing cost pool, against about 28% for the cutting itself. That pool is a processing cost and does not include the wood, so it says nothing about a primary mill, where the log is the bigger number.

The data explains part of the lag. Setup time lives in production logs, sometimes on paper. The sequence of runs lives in the schedule, and the orders that caused each changeover live in the ERP. None of them is in a camera's field of view, and a schedule that accounts for setups needs all three read together, over weeks. A first check needs no AI at all. Take a quarter's production log, count the changeovers on one line, and set the hours they took beside the orders that caused them. If a handful of short orders account for most of the changeovers, the schedule is where the money is.

How to tell where AI will pay

Before adding AI to a line, split its processing cost into cutting, setup and changeover, maintenance and waiting, and set the value of the wood beside it. In a primary mill, the wood usually dominates, and the proven AI on the saw and the grader is the right place to start. In a remanufacturing or millwork line with many short runs, if setups are the larger share, the first gain may be in the schedule.

When it doesn't apply

High-volume commodity lines that run one product for long periods have few changeovers, and for them grading and optimisation are the right focus. The cost split above comes from one manufacturer's remanufacturing pool. Other operations will split differently, which is the reason to measure your own before choosing.

Quarri for wood products is built around how a wood products plant runs, where the order book meets real capacity.

Quarri is an AI-native data platform for the timber supply chain. It connects buying, production, sales and inventory for forest management, sawmill, wood products and pulp, paper and packaging operators.