For decades, grading a board meant a trained eye walking the green chain, calling stud, common or select by feel and sight, one piece at a time. That is changing fast at the world's largest sawmills, where camera arrays, X-ray, 3D laser and deep neural networks now do the calling — faster, and more consistently, than any human crew ever could. For buyers sourcing timber internationally, the shift quietly resets what "consistent grade" is supposed to mean, whether or not the mill on the other end of the order has made the jump yet.
- 240 boards/minute — the speed at which Finnos' newest board scanner reads all four faces of a board without rotating it, grading at over 96% accuracy (Wood & Panel Europe, July 2026).
- 250+ systems — AI-enhanced optimization systems BID Group has running in sawmills worldwide; USNR runs over 100 systems on deep neural networks after a decade installing AI grading technology (Forestnet).
- 2006 → 2018 — machine-learning grading systems entered sawmills in 2006; deep-neural-network systems followed in 2018, and adoption has accelerated since (Forestnet).
- 100 boards/minute — the scanning speed of USNR's latest hardwood grading system, trained to flag knots, checks, splits, bark pockets, decay and insect holes against NHLA grading rules (Miller Wood Trade Publications).
From a Trained Eye to a Trained Network
Machine-learning-based grading isn't new — it has been showing up on sawmill floors since 2006, mostly from equipment makers like Comact and Autolog automating primary breakdown and trimmer optimization. What changed is the model underneath it. Starting around 2018, BID Group and USNR began shipping systems built on deep neural networks instead of rule-based image processing, and the difference shows in the numbers: BID Group now has more than 250 AI-enhanced optimization systems operating in the field, and USNR — a decade into installing AI grading technology — runs over 100 systems with deep neural networks doing the classification (Forestnet). Neither company treats this as experimental technology anymore. It is standard equipment on new production lines.
What the Newest Scanners Actually Catch
The current generation goes well beyond spotting a knot. Finnos' Fusion platform combines X-ray imaging, 3D laser measurement and RGB cameras in a single pass, and its Board Scanner reads all four faces of a board without rotating it — up to 240 boards a minute, grading at better than 96% accuracy. Its Neuro CT system takes the concept further: a CT scan grades the log virtually before it is even sawn, so the mill can choose the cutting pattern that yields the most value from that specific log rather than a generic average (Wood & Panel Europe, July 2026). On the hardwood side, USNR recently walked members of the Appalachian Lumbermen's Club through a system that scans up to 100 boards a minute, using a deep neural network trained to flag knots, checks, splits, bark pockets, decay and insect holes, preloaded with NHLA grading rules out of the box (Miller Wood Trade Publications).
The Adoption Gap Is Real — and It Isn't Closing Evenly
This capex is concentrated where mills have the volume to amortize it: large softwood and hardwood producers in North America and Northern Europe running high-throughput lines. Forestnet's own read on where this goes puts a number on it — within a decade, most modern sawmills will need some version of this technology just to stay competitive. But "modern" in that sentence means mills already committed to automation, not the plantation sawmills across Misiones, Corrientes and Uruguay that supply most of the world's export pine and eucalyptus. Those mills still grade visually, board by board, trained crew against a rulebook — which isn't a shortcoming, it's simply the standard everywhere outside the small group of operations at the top of the capex curve.
What This Means for Buyers Sourcing From South America
As AI-graded consistency becomes something buyers routinely experience from Nordic and North American suppliers, it quietly raises the bar for what "consistent grade" is expected to mean — including from mills that have no scanner on the floor. Until machine grading becomes standard across export-grade plantation sawmills, and it is years away from that in South America, the reliability a buyer actually gets depends less on a price list and more on how rigorously the trading partner behind it vetted the mill's real grading process. That is the practical argument for in-person sawmill visits before any supply relationship starts — confirming who grades the wood, against which rules, and with what documentation, rather than assuming "KD" or "APG" means the same thing from one mill to the next.
For a buyer weighing a new South American supplier, the useful question isn't whether that mill runs AI scanners — almost none in the region do yet. It's how the mill grades without one, and how much of that process the intermediary placing the order has actually seen firsthand.