Can AI optimise purchase orders across suppliers?

Yes, where there is a choice to optimise and the data compares. For items with two or more qualified suppliers, a model can weigh price, actual lead time, on-time record, minimum quantities and freight, and recommend who to order from and how much. The first limit is usually the data rather than the algorithm. At one lumber and millwork manufacturer, most stocked items had only one supplier on record, and where there are several, prices, units and lead times are often held in forms that can't be compared. Building that data comes first. It also shows which items need a second source, and which are single-sourced on purpose.

What the usual answer says

The usual answer describes AI comparing suppliers on price, lead time, reliability and order minimums, and recommending the best supplier and quantity automatically. That is what an optimiser does once it has options and comparable data. The answer takes both for granted.

How many items have a choice

From Quarri's own work with a lumber and millwork manufacturer: about 7 in 10 stocked items had a single supplier.

For those items, the optimisation question is when and how much to order, not from whom. That is one business, and we know of no industry figure for lumber.

The data problem is wider. In the 2026 CPO Report from ProcureAbility, which sells procurement services, and ProcureCon, 54% of procurement leaders "reported insufficient data quality and cross-system integration as limiting factors" for AI. Only 11% said they were "fully ready".

What the research can and can't show

Research on dual sourcing is about the ordering policy once two options exist. Fabian Akkerman and colleagues, in a study first published in October 2024, tested ordering policies for spare parts in the energy sector, where each part could come from conventional or additive manufacturing. Their policies "outperform the baseline in 91.1% of instances, yielding average cost savings up to 22.6%". That is a better policy against a simpler one, both with two options. It shows what good ordering is worth when the choice exists, not what a second source is worth.

Melvin Drent and Joachim Arts, in a 2022 paper, model a regular supplier and a faster, dearer one: "The expedited supplier has a shorter lead time than the regular supplier but charges a higher unit price". The value of the second supplier depends on that trade-off and on the cost of running short, and neither can be calculated without actual lead times and prices for both. Both papers are about models, not lumber yards.

Why supplier data rarely compares

A purchasing system that holds the last price paid to each supplier in whatever unit it was invoiced can't compare suppliers directly, and promised lead times differ from actual ones. So the work before any optimisation is data work: match item descriptions across suppliers, convert prices to one unit per item, record actual lead time from order date to receipt, and record whether each delivery was complete and on time.

Timing for the single-sourced

For single-sourced items, the optimisation that remains is timing and quantity. Ordering against actual lead times rather than promised ones, grouping items from one supplier to meet freight breaks and minimums, and moving reorder points when a supplier's recent record slips all need the actual lead-time data above, and no second source.

When concentration is the point

Some single-sourcing is a choice. Concentrating volume with one mill can buy programme pricing, rebates and allocation when supply is tight. An optimiser that splits orders per purchase can win a few cents on one order and lose the terms that came with the volume. So the recommendation should show the trade-off: the saving on this order against the effect on volume commitments. For single-sourced items with high spend or long lead times, the data can make the case for qualifying a second supplier, and that remains a commercial decision.

When it doesn't apply

Items bought under a fixed contract with one supplier, by design, have no choice to optimise. Very small businesses buying a few items from one yard may gain more from better timing than from supplier choice. And for specialty items where only one mill makes the product, single-sourcing is the market, not a gap.

Quarri for operations teams is built for the people running the crews, the lines and the yard.

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.