October 29, 2026
Fashion uses AI every day. It does not trust it yet.
AI is already part of the working day in fashion. Trusting it with decisions that matter is another thing. Here is why the gap exists and what closes it.
The short answer
Around nine in ten fashion professionals now use AI every day, but only about a quarter are willing to base important decisions on it, according to The Interline's 2026 survey. The gap persists because most AI sits outside the systems that hold product truth. Trust follows when AI works from governed product data and every output can be traced back to its source.
Key takeaways
- Usage is nearly universal; trust for important decisions is not: about a quarter, per The Interline.
- The gap is a loop: AI is kept away from PLM and ERP because it is not trusted, so it never gets the grounding it needs to earn trust.
- Traceability beats confidence scores. People trust an answer they can check against the record it came from.
- Close the gap in small steps: ground AI in one product record, keep a person approving, and widen its job as it proves itself.
What do the numbers actually say?
The headline from The Interline's 2026 survey of fashion professionals, written up in Arm's Length AI is a contradiction. Around nine in ten people working in fashion use AI every day. Only a quarter or so would peg an important decision to what it produces. Roughly three in five worry that weak or invented outputs will lead to poor choices.
Those numbers describe a tool that is useful for drafting, summarising and exploring, and not yet relied on where money, margin or reputation is at stake. That is a reasonable position for professionals to take. A wrong colour on a mood board costs nothing. A wrong measurement in a tech pack costs a sample round, and a wrong price on a line sheet costs a customer relationship.
| Finding (The Interline, 2026) | Share of professionals |
|---|---|
| Use AI every day | About nine in ten |
| Willing to base important decisions on AI | About a quarter |
| Fear sub-optimal choices from inaccurate outputs | About three in five |
| Say their organisation expanded AI activity in the past year | About three quarters |
Why does the gap persist?
The Interline describes a loop that we recognise from conversations with product teams. Because people do not trust AI, it is kept at arm's length from the systems that hold product truth: PLM, ERP, the approved style. Because it is kept at arm's length, it works from whatever someone pastes into a chat window, which makes its answers less reliable, which confirms the distrust.
The loop is not really about the model. A general-purpose model asked about a style it cannot see will guess. The same model working from the approved version, the bill of materials and the comment history has far less room to guess. Most of the trust problem is a context problem.
What makes an AI output trustworthy?
On The Interline's podcast How Much Will Fashion Let AI Decide?, Rupert Schiessl of Bamboo Rose put the point plainly: “You can’t own something you can’t explain.” He argued that trust comes from being able to trace an output back to the records it was built from, rather than from a confidence score attached to it.
We think that is the right test, and it is a practical one. For any AI output a product team is asked to rely on, three questions should have quick answers:
- Where did this come from? Which style, which version, which file or data.
- What did it not know? Missing values should be flagged, not filled in with a plausible guess.
- Who approved it? A named person signed it off before it left the team.
How do teams close the gap without a big-bang rollout?
Trust is earned job by job. The teams we see making progress pick one piece of work where the inputs are clear and the output is easy to check, and widen from there.
1. Pick a checkable job
Something with a right answer your team can verify, such as a line sheet built from approved styles, or a tech-pack draft built from drawings and a BOM.
2. Ground it in one record
Give the AI the approved style and its data, not a copy pasted into a chat. One source means one version of the truth.
3. Keep a person approving
Nothing ships without sign-off. The approval is recorded on the work, so the history explains itself.
4. Measure the misses
Count what reviewers had to fix. When the misses drop, give the AI a slightly bigger job.
What does this mean for product teams?
The trust gap is not a reason to slow down. It is a sign that the next gains will come from connecting AI to product data, not from adding more standalone tools. Every tool that works outside the style record adds another place where the truth can drift.
It also means the most valuable AI in fashion may look modest. It drafts, assembles and checks against the record, and it says clearly what it does not know. That is less dramatic than a perfect image from a prompt, and it is what people will actually trust with decisions.
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 mid-sized womenswear label uses AI in three places: designers generate mood boards, the e-commerce team writes product copy, and a merchandiser asks a chat assistant to summarise last season's sell-through from a spreadsheet she pastes in. Nobody would let the assistant set prices.
The team picks one checkable job to ground: building the wholesale line sheet from approved styles. The AI reads the approved style records directly, not a pasted copy, and marks three styles with no wholesale price instead of inventing one. The merchandiser fills the gaps, approves the sheet and sends it.
Over two seasons the reviewers stop finding errors, so the team gives the same approach a bigger job: buyer-specific decks. Trust grew because every answer could be checked against the record it came from, and because the system said plainly what it did not know.
Where CotonAI fits.
CotonAI is built on two bets that follow from this gap. First, one style record that every team and document builds from beats isolated tools: renders, colourways, imagery, line sheets and tech-pack drafts all come from the same approved version. Second, agents do the work and people approve: the CotonAI Agent does delegated work with the brand's context and brings it back for sign-off, and approvals are recorded on the style.
Missing prices, measurements and trims are flagged, not guessed. Your work is never used to train AI models, and every access to your workspace is logged. See the facts.
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.
- The Interline, Arm's Length AI, Ben Hanson, 18 August 2026. Essay drawing on The Interline's 2026 AI survey
- The Interline, How Much Will Fashion Let AI Decide?, Ben Hanson with Rupert Schiessl of Bamboo Rose, 1 October 2026. Podcast
Common questions
Do fashion professionals trust AI?
Most use it daily, but only about a quarter are willing to base important decisions on it, according to The Interline's 2026 survey.
Why do people distrust AI outputs in fashion?
Mostly because AI works without the product context it needs, so it can produce plausible but wrong answers. About three in five professionals worry about this, per The Interline.
How can a brand build trust in AI?
Ground AI in the approved product record, make every output traceable to its source, flag missing data instead of guessing, and keep a person approving before anything ships.
Is a confidence score enough to trust an AI answer?
Not on its own. Being able to see which records an answer came from, and who approved it, is a stronger basis for trust.
Keep reading
Ground your AI in the style record. Then trust it with more. We'll show you the workflow end to end and recommend where to start. Book a demo