AI inside PLM: what did the major fashion PLM vendors add in 2025 and 2026?

In 2025 and 2026, the main fashion PLM vendors added three kinds of AI: extraction that turns sketches and documents into structured tech pack data (PTC FlexPLM), generative studios for sketches, colourways and imagery tied to PLM records (Centric AI Studio), and agentic platforms that automate parts of product development end to end (Lectra Apogy). The shared idea is to bring AI inside the system of record, so outputs carry product data and approvals with them. Independent evidence of results is still thin.

What did the major vendors announce?

The table summarises the announcements we could verify in primary or trade sources. It is not an exhaustive market overview, and inclusion is not a recommendation.

How does AI tech pack extraction work in PLM?

PTC describes its FlexPLM addition as automating tech pack creation: the system extracts data from design drawings and populates bills of materials, measurements, construction details, attributes and colourways. PTC positions this as removing manual data entry at the handoff between design and development, reducing errors and rework and shortening the path from concept to production ready sample.

The idea addresses a real bottleneck. Product developers spend a large share of their time transferring information from sketches and emails into structured fields. The open questions are accuracy on complex garments, how uncertain values are flagged, and whether extracted data is clearly marked until a developer confirms it.

Read also

What does 3D sampling really save? Sample costs, lead times and carbon

What are generative design studios inside PLM?

Centric Software launched Centric AI Studio as a generative AI platform natively integrated with Centric PLM. According to the announcement, it generates sketches, concepts, colourways and product variants, creates commerce ready imagery with less dependency on photo shoots, and keeps outputs connected to live PLM product data and approval workflows. Centric's chief AI officer framed the goal as moving brands beyond "disconnected experimentation".

That framing is the most significant shift of the period. Many design teams already used general image generators in 2024 and 2025, but the outputs lived in folders outside any system, with no link to the style record, no version history and unclear rights. Embedding generation in PLM makes outputs traceable, which matters for approvals and for later questions about how an image or design was produced.

3D is the bridge between both approaches. Integrations such as the one announced by DeSL and Browzwear in January 2025 let teams visualise colour and material variations in 3D and annotate 3D models in real time with vendors, inside the PLM workflow. Generative and agentic tools increasingly build on the same digital garments and material libraries, so brands with an established 3D practice and a maintained digital material library are better placed to use the new AI features than those starting from flat sketches alone.

What does agentic AI mean for product development?

Lectra's Apogy goes furthest in its positioning. FashionUnited reports that it combines AI agents, product data and design processes in one cloud environment, generates design concepts from prompts, sketches or images, supports 3D simulation and rendering, and automates routine tasks. Early adopters named are O'Neills in Ireland and Oniverse in Italy. FashionNetwork reports that around 30 brands and manufacturers trialled the first version, and that Lectra suggests luxury companies could reduce product development time by 30 percent while fashion companies could double or even triple the number of products developed.

Two details deserve attention. First, FashionNetwork reports that Lectra earns revenue on the number of product developments completed rather than on user licences, a model that aligns vendor revenue with usage. Second, the performance numbers are forward looking claims, not audited outcomes.

What is still missing?

• Independent evidence: we found no published, independently measured results for the 2026 launches yet; most figures come from vendors or their customers.
• Data prerequisites: extraction and generation depend on clean material libraries, consistent block patterns and well maintained value lists. Weak master data limits all three tool types.
• Fit and construction accuracy: a convincing render is not proof that a garment can be made as shown. Technical validation still needs pattern makers and fit sessions.
• Governance: rights to training data and generated designs, and labelling of AI-generated imagery, remain questions for legal and brand teams.
• Cost transparency: usage based pricing can be attractive, but needs modelling against realistic development volumes.

Read also

How do you make 3D the default in PLM? Workflow, approvals and file standards

How should brands evaluate AI features in PLM?

1. Pick one painful process (for example tech pack creation for a core category) and measure today's time and error rate.
2. Run the AI feature on real styles from a past season, where the correct answer is known.
3. Measure accuracy per field, not overall impressions, and record how often a person had to correct the output.
4. Check how machine generated values are flagged, versioned and approved in the PLM.
5. Estimate total cost including data clean-up, training and change management, not only licence or usage fees.

The direction of travel is clear: AI is moving from side tools into the core product record. Whether it pays off depends less on which vendor a brand chooses and more on the quality of the product data the AI starts from.