Every promise about AI agents in fashion runs into the same wall: scattered, stale product data. Fixing it is unglamorous, and it is the work that makes agents useful.

The short answer

AI agents in fashion can only act on product data they can reach and trust. Most brands keep that data in islands: PLM, ERP, spreadsheets, email and supplier files. Before deploying agents, teams need one governed source of product truth, with permissions in the data and a named owner for every agent. As one factory-software leader told Apparel Resources, nothing changes until the data foundation is fixed.

Key takeaways

  • Agents are only as reliable as the product data they can see.
  • Product data usually lives in islands (PLM, ERP, spreadsheets, email), from design office to factory floor.
  • Regulation, including the EU's Digital Product Passport, is adding pressure to keep product data complete and current.
  • Start with one record per style that every team and document builds from.

Why is product data the bottleneck?

Ask an AI agent which styles share a fabric affected by a new tariff, or which colourways still lack an approved price, and its answer depends entirely on whether it can see the right records. In most fashion businesses those records are spread across PLM, ERP, spreadsheets, email threads and supplier files, each slightly different.

The problem runs all the way to the factory. Apparel Resources described factories running planning, production and maintenance as separate islands, and quoted Anuj George Kurian of Solvei8: “Everyone asks for visibility, but nothing changes until the data foundation is fixed.”

Why is PLM back in focus?

The Interline's 2026 spotlight on PLM called product lifecycle management, by its estimation, the most mature technology category in fashion, and argued that AI demand is bringing it back to prominence because copilots and agents belong where product data lives. That instalment is sponsor-supported by PLM vendors, so the enthusiasm is expected, but the logic holds regardless of which system holds the data: agents need a governed source to work from.

On another Interline conversation, Rupert Schiessl argued that permissions belong in the data layer rather than in each application, and that every agent should be attached to a responsible human. Both points are about data governance, not models.

How does regulation raise the stakes?

The EU's Ecodesign for Sustainable Products Regulation introduces a Digital Product Passport, a digital record of product information intended to travel with the product, and textiles are among the product groups the Commission has prioritised. The details for each product group are set in later rules, so check the Commission's pages for current timelines rather than relying on any summary, including this one.

In an Interline interview, a sustainability-data vendor argued that every sustainability programme rests on clean, fast data ingestion, and that integrating scattered systems is often the slowest part. Whatever the exact rules end up requiring, brands with complete, current product records will find compliance far easier than those reconstructing data from email.

What does a good product-data foundation look like?

Where should a brand start?

  1. 1. Pick one category or season

    Not the whole business. A bounded set of styles is enough to prove the approach.

  2. 2. Make one record per style the source

    Every document and image is built from it, not alongside it.

  3. 3. Flag the gaps

    Missing prices, measurements, materials and certifications become visible tasks.

  4. 4. Connect, do not copy

    Sync with the systems that already exist instead of re-entering data.

  5. 5. Then add an agent

    Give it one owner, one job and access only to what it needs.

What does this look like in practice?

An illustrative example. The brand and the details are invented to show the idea; they are not a case study.

A brand wants an agent to answer one question for buyers: which styles in the spring range are certified organic. Today the answer lives in a PLM field that is half filled in, a sustainability spreadsheet, and supplier emails with certificates attached.

Instead of starting with the agent, the team starts with the record. For one category, it makes the style record the single place where the material, its certificate and the certificate's expiry date live, and flags every style where they are missing. Only then does it add the agent, owned by the sustainability lead, with access to that category alone. The agent's answers are now traceable to a certificate on a record, which is what the buyer needed.

Where CotonAI fits.

One style record that every team and document builds from is the core of how CotonAI works: designs, renders, colourways, materials, images, line sheets and tech-pack drafts all hang off the same style, with versions and approvals recorded. Missing values are flagged rather than guessed. CotonAI has 100s of integrations across storefronts, marketplaces, PLM, ERP and ads, so it can sit alongside existing systems. It does not produce Digital Product Passports. See integrations.

Sources.

We read these pieces and wrote this analysis in our own words. Figures are credited to the source each article gives; vendor and company figures are their own claims.

  1. Apparel Resources, From Fabric to Factory Floor, Fashion's Digital Shift is Going Full Throttle, Dheeraj Sharma, 22 December 2025.
  2. The Interline, Spotlight On: PLM, Data Management, AI & Collaboration, 26 October 2026. Sponsor-supported instalment
  3. The Interline, How Much Will Fashion Let AI Decide?, Ben Hanson with Rupert Schiessl of Bamboo Rose, 1 October 2026. Podcast
  4. The Interline, All About AI: Atnyel Guedj of Made2Flow, 15 September 2026. Vendor interview
  5. European Commission, Ecodesign for Sustainable Products Regulation, accessed 29 October 2026. Official page

Common questions

Why do AI agents need clean product data?

Because they can only act on what they can see. Scattered or stale data leads to wrong answers; one governed record per style gives them a reliable source.

Is PLM still relevant with AI?

Yes. The Interline argued in 2026 that AI demand is bringing PLM back into focus, because agents need a governed home for product data.

What is the Digital Product Passport?

A digital record of product information introduced under the EU's Ecodesign for Sustainable Products Regulation. Textiles are among the prioritised product groups; check the European Commission for current timelines.

Where should permissions for AI agents live?

In the data layer, so an agent can only see and change what its human owner is allowed to.

Keep reading

Fix the record. Then add the agent. We'll show you the workflow end to end and recommend where to start. Book a demo