AI-assisted bill of materials and costing from a design brief
How AI can pre-populate a bill of materials and estimate labour and cost from images and history, where it needs human verification, and how to set it up sensibly.
KEY TAKEAWAYS Summary by the editors
- AI can suggest bill of materials components from historical data for similar styles, but new materials and new suppliers still require human review, according to a 2026 PLM vendor blog.
- A Coats Digital announcement describes GSDQuest, an AI tool that builds a Bill of Labour from as little as one garment image using standard minute values, a labour estimate that is separate from material cost.
- The same PLM vendor blog reports that early adopters see 40 to 70% less first-draft time on reorders and carry-forwards, but this figure comes from unnamed vendor case studies, not independent research.
- A 2025 Drexel University thesis found that vision models extract garment features from images robustly, which is the capability that image-based costing and BOM tools depend on.
- Costing from a brief is an estimate: factory-specific notes, tolerances, supplier quotes and construction decisions remain human-authored.
AI-assisted costing from a design brief works by recognising what a garment is made of and how it is constructed, then matching that to historical bills of materials, labour times and supplier data to produce a first estimate. It is a fast starting point for early design decisions, not a quote. Material callouts, supplier prices and factory-specific details still need people to verify them.
What does a bill of materials and costing involve?
A bill of materials lists every component of a style: fabrics, trims, linings, labels, packaging, with quantities and suppliers. Costing builds on it by adding material prices, labour, overheads, freight, duty and margin. Early in design, many of those inputs are unknown. Teams therefore estimate from comparable past styles, and that is where AI tools claim to help.
What can AI automate in a BOM today?
A 2026 blog by PLM vendor Wave PLM, which is promotional and should be read as such, lists what it says is working in practice. AI can suggest BOM components from historical data for similar styles and auto-populate supplier codes. It can pre-populate a new style to 60 to 70% completion from the closest historical match, and it can suggest standard grading increments and flag missing measurements. The blog also gives the limits: measurement values need full review, material callouts need sourcing verification because the AI lacks supplier knowledge, and new materials or suppliers require review. The blog says construction notes can be suggested in standard language but not originated by AI, and that factory-specific notes and tolerances remain human-authored.
The blog reports that early adopters see 40 to 70% less first-draft time for reorders and carry-forwards, based on vendor case studies from 2024 to 2025. Because the studies are unnamed and from a vendor, treat the figure as a claim to test in your own pilot rather than a benchmark.
How can costing be estimated from an image or brief?
Coats Digital announced GSDQuest, an AI tool that extends its GSDCost solution. According to its release, the tool can generate a standardised Bill of Labour from as little as one garment image, by analysing images to identify design and construction elements and mapping them to the company's QED Library. The release says it builds on a time-motion methodology using internationally recognised standard motion codes and standard minute values, and works with images, PDFs and tech packs. The company says the tool reduces costing from hours to seconds. This is a vendor statement, and the release gives no independent evaluation of accuracy.
The underlying idea, recognising garment features from pictures, is supported by academic work. The 2025 Drexel thesis on pattern automation found that its vision component performed robustly in extracting garment features, even though the downstream pattern generation struggled with complex garments. Recognition is therefore the more mature half of the problem. Turning recognised features into a trustworthy cost still depends on the quality of the underlying time and material data.
What is the workflow from brief to costed BOM?
- Write the brief with category, fabric type, key construction details and target cost.
- Attach reference images or a technical flat.
- Let the tool propose a closest historical match and a draft BOM.
- Review each material line, replacing placeholders with approved or quoted materials.
- Estimate labour from standard minute values or factory data and compare with the tool's output.
- Request supplier quotes for new items and update the cost sheet.
- Approve the costing, and record which values were suggested, edited or confirmed.
| Element | Typical AI contribution | Verification needed |
|---|---|---|
| Components and trims | Suggested from similar styles | Confirm against the actual design and approved library |
| Supplier codes | Auto-populated from history | Check validity, minimum order quantities and lead times |
| Fabric consumption | Not established in the sources cited | Marker-based or measured consumption |
| Labour minutes | Estimated from images using standard minute values (vendor claim) | Compare with factory time studies |
| Material prices | Historical values | Current supplier quotes |
| Construction notes and tolerances | Standard language only | Human-authored, factory-specific |
What data do you need first?
These tools depend on structured history. A system can only match a new style to a close precedent if past styles have complete, consistent BOMs, accurate final costs and clear category labels. The Wave PLM blog lists connecting the tech pack to BOM, costing and sample records without manual re-entry as an evaluation criterion, which hints at the real prerequisite: one connected record per style. Companies with BOMs in spreadsheets and email attachments should expect to clean data before any AI feature works well.
Estimated labour needs its own data. The Coats Digital approach relies on standard minute values; a company using other methods must check how its factories measure and quote operation times.
Where are the risks?
- False precision: a tool returns a precise number from an imprecise input, and that number is anchored in negotiations.
- Stale history: past prices and suppliers no longer reflect today's market.
- Hidden assumptions: the user cannot see why a component was suggested.
- Vendor lock-in: costing logic and libraries sit in a proprietary system.
- Confidentiality: images, briefs and supplier data sent to external tools need contractual protection.
Finally, decide who owns the output. Costing sits between design, sourcing, product development and finance, and an AI estimate can slip into a negotiation without anyone having approved it. A simple rule helps: any AI-generated figure is labelled as an estimate in the record until a named person confirms it against supplier or factory data.
A short pilot gives the best evidence. Take twenty closed styles from a recent season, run them through the tool as if new, and compare the suggested BOM and labour estimate with the final approved versions. Record how many lines were right, how many needed edits and how long the review took. That comparison, rather than a vendor figure, tells you whether the tool earns its place in the process.
Frequently asked questions
Can AI create a bill of materials from a design brief?
It can suggest one. According to a 2026 PLM vendor blog, AI can propose components and supplier codes from historical data for similar styles, but new materials and suppliers require review and measurements need full human checking.
How does AI estimate garment cost from an image?
Coats Digital's GSDQuest, as described in its release, identifies design and construction elements in images and maps them to a library of standard motion codes and standard minute values to build a Bill of Labour. This covers labour only and is a vendor claim without published accuracy figures.
How accurate is AI costing?
The sources cited give no independent accuracy figures. Vendor reports of 40 to 70% less first-draft time relate to time saved on tech packs for reorders and carry-forwards, not to cost accuracy. Test any tool against your own closed styles.
What data is needed for AI-assisted costing?
Complete, consistent historical BOMs, final costs, supplier data and category labels, ideally in one connected record with the tech pack. Labour estimates also need standard time data or factory time studies.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
