Which questions can AI answer for a millwork manufacturer?

AI can answer two kinds of questions for a millwork manufacturer. The first kind is about documents: what an architectural drawing calls for, what a shop drawing should contain, what a customer is asking. That is where the millwork AI tools in the search results sit. The second kind is about the business's own records: which customers are buying less, which stocked items have stopped selling, what a product costs once setups are counted. These are the questions that change decisions, and AI can answer them only when sales, purchasing and stock records are joined.

What the usual answer says

The top search results sell the first kind. Tools for estimating and drafting read drawings and produce takeoffs, shop drawings and cut lists. An AI assistant for millwork shops offers to "Answer frequently asked questions instantly" and "Qualify leads before they reach your team". Grant Thornton's August 2026 list of AI questions for manufacturers sorts uses under throughput, quality and cost. Its partner Sumeet Mahajan puts the test as "How will this make us more resilient?"

Those uses are real, and for a shop whose bottleneck is quoting, an estimating tool may be the right first purchase. None of these pages addresses the questions a millwork owner asks about the business itself.

What small manufacturers are up against

The OECD's December 2025 report on AI adoption by small firms puts the gap in numbers. Across member countries, 40% of large firms use AI, against 11.9% of firms with 10 to 49 employees. The share of firms using AI rose from 5.6% in 2020 to 14% in 2024. Manufacturing grew more slowly than the leading sectors, with adoption up by about 70% over 2021 to 2024.

On the obstacles, the report cites a 2025 study it produced with BCG and INSEAD. For small firms, "especially those in the manufacturing sector", these include "difficulty finding vendors of AI solutions tailored to their needs, lack of quality data and digital readiness". In our reading, the first may explain why millwork AI clusters around drawings, where one tool can serve many shops. The second may explain why the record questions go unanswered, since they need a shop's own data in a state a model can use.

The questions records can answer

Record questions share a shape. The headline answer is easy to get and looks fine, and the answer broken into its parts says something different. Revenue by year is in any accounting system. Revenue split into what existing customers spent this year against last year, and what new customers added, needs every invoice line tied to a customer across two years.

From Quarri's own work with a lumber and millwork manufacturer: revenue grew 25%+ in a year while net revenue retention was about three quarters: losses and shrinking accounts were hidden by new business.

Other questions have the same shape. Stock on hand is a total, and stock that has not moved in a year is a different list. A standard cost is one figure, and the cost including set-ups can be another. A supplier's on-time record can look fine in aggregate and different order by order. An AI system can surface each of these, and each needs two records joined: invoices and customers over time, stock and sales, production logs and costing, purchase orders and receipts.

Where the two kinds meet

For a shop that works project by project, the most useful record question starts from a document. Estimate against actual by job compares what the estimate said a job would take, in material, hours and machine time, with what the job records show it took. The estimate is a document; the actuals are records from purchasing, time sheets and production. AI can read the estimate into lines and match each to the actuals, job after job. The pattern across jobs, such as which product types or customers are always underestimated, is what feeds the next quote.

A test for any AI tool

Pick one question whose headline and parts could disagree. Revenue growth split into existing and new customers is a good one, because the headline is easy to get and the split usually isn't. Ask the tool, or the person who would build the answer, how long it takes and what records it reads. If the answer arrives in minutes and ties to the accounts, the records are joined. If it takes a week of spreadsheets, the first job for AI is the join, and a chatbot on top of the current systems will answer the headline only.

When it doesn't apply

A shop doing one-off architectural projects with few repeat customers has little use for retention figures; for it, estimate against actual is the record question that matters. A very small shop whose owner knows every customer and every rack gets less from decomposition, because the owner already carries it. And where quotes are the constraint on growth, estimating comes first.

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.