Designer and LCI Melbourne’s business and fashion industry expert Natasa Pitra-Grbic reveals why AI is more valuable in fashion production
Fashion’s AI conversation has been dominated by what the customer can see: campaign images, copywriting, social media and personalisation. These tools are visible, fast and seductive. But the most commercially valuable application may be far less photogenic. It sits behind the scenes in sourcing, product development, costing, range planning and production.
If a social caption underperforms, it can be rewritten tomorrow. If a range is overbought, a landed cost is miscalculated or a production approval is late, cash can be tied up for months. That is why I often say that production mistakes cost more than marketing mistakes.
AI can accelerate the work around a supplier relationship, but it cannot build the trust on which production depends.
This matters in Australia now. The Westpac–Melbourne Institute Consumer Sentiment Index rose to 83.9 in July, yet remained in the bottom 10 per cent of readings across its 50 year history. The Reserve Bank’s August outlook also pointed to elevated inflation, higher input costs and moderating domestic demand. Meanwhile, only 9 per cent of retail businesses used AI in 2024–25, according to Australian Bureau of Statistics data reported by Ragtrader. With cautious consumers and fragile margins, every product decision has to work harder.
AI is not one tool it is a system.
It helps to separate three capabilities often bundled together. Generative AI can organise unstructured information, review documents and prepare a first draft. Predictive AI can analyse historical sales, inventory and demand patterns. Workflow automation can move information between systems, trigger tasks and flag exceptions. A chatbot is not a product lifecycle management system, and a prediction is not a purchase order. Value comes from matching the right capability to a clearly defined production decision, supported by clean data and a human approval point.
1. Turn supplier research into a decision process
Supplier research is rarely neat. One factory replies by email, another through a spreadsheet and another through a messaging app. AI can organise minimums, payment terms, lead times, capabilities and certifications into a consistent comparison, surface missing answers and prepare follow up questions. It can help a business compare factories against its own criteria, rather than simply ranking whoever appears first in a search.
At the Global Sourcing Expo Sydney this year, I demonstrated how supplier information could be structured around minimums, lead time, manufacturing capability, communication, sample quality and risk. The most valuable output was not a machine generated recommendation. It was a clearer view of what still needed to be verified by a person. A slower factory quote, for example, may reflect careful assessment of construction and labour rather than poor capability. Certifications still need to be checked, samples tested and terms confirmed in writing. Humans must negotiate, read the room and solve problems. AI makes the administration faster; it does not create trust.
2. Make costing and range planning dynamic
A quoted unit price is not a landed cost. The true figure may include development, sampling, fabric, trims, testing, quality control, packaging, freight, duty, currency movement, warehousing and payment fees. AI can help teams model what happens when quantity changes, freight rises, the exchange rate moves or another sample is required. It can flag blank inputs, duplicated costs and the financial difference between air freight and a delayed launch. This is where cash is protected.
The same approach can strengthen range planning. AI can compare sell through, margin, returns, size curves, colour performance and channel, then test different mixes of hero, core and fashion products. For smaller brands, the first benefit may be discipline rather than prediction: exposing too many colours, duplicated price points or more stock keeping units than the budget can support. The human still interprets the result. A slow seller may have launched late, been photographed poorly or reached the wrong channel. Data shows what happened, not always why.
3. Create visibility before a problem becomes expensive
Fashion production is a chain of dependencies: materials, technical packs, samples, fit comments, testing, purchase orders, cutting, sewing, quality control, freight and launch. AI can review a technical pack for missing measurements or unclear tolerances, turn a launch date into a backwards calendar and show which approvals sit on the critical path. It can also maintain a risk register covering material delays, factory shutdowns, weather events, shipping disruption and public holidays.
This is where workflow and administration change most visibly. Work that once took hours of reconciling emails, spreadsheets and meeting notes can begin with a structured first pass in minutes. The production coordinator then verifies the information, speaks with suppliers and decides what action to take. The speed comes from reducing manual assembly, not removing accountability.
AI is a margin tool before it is a headcount tool.
For businesses facing cautious consumers and rising costs, the strongest case for AI is not simply doing the same work with fewer people. It is avoiding expensive outcomes: excess stock, repeated samples, rush freight, missed delivery windows, quality claims and markdowns. The sensible starting point is one recurring decision with a measurable cost, such as supplier comparison, landed costing or production timeline management. Define the inputs, expected output and human approval point, then measure whether the decision improves.
The most common mistake is buying a tool before understanding the process. AI will scale a sound workflow, but it can also accelerate confusion when ownership, data and approvals are unclear. Businesses also need governance. Supplier pricing, contracts, employee information and unreleased product data should not be pasted into an unapproved public tool. Sources must be verified, access controlled and every consequential recommendation reviewed by a named person.
What will fashion teams look like?
The future team is not a room without people. It is a team in which specialists spend less time formatting, copying, chasing and reconciling, and more time negotiating, developing suppliers, assessing quality, resolving fit and making commercial trade offs. Product developers will use AI to check information and surface exceptions, but they will still need to understand construction, patternmaking, fabrication and how a garment behaves on the body. Sourcing professionals will combine relationship building with faster evidence based comparison.
Some teams may become leaner, but removing junior roles too aggressively creates an industry with tools and no pathway for people to develop expertise. AI literacy should sit beside production literacy, not replace it. Larger organisations may add workflow owners and supply chain data specialists; in smaller businesses one person may wear several hats. In both cases, every automated output needs an accountable human owner.
The opportunity is behind the scenes.
The biggest AI opportunity in fashion may never appear on the Instagram grid. It may look like one fewer sample round, one avoided air freight bill, a clearer supplier brief, a smaller and more profitable range or a risk identified three weeks earlier. Content helps us sell the product; production determines whether the product is worth selling. AI will not replace relationships, craftsmanship or judgement. Used properly, it gives experienced people better workflows, stronger systems and less administration so they have more time for the human work on which production depends.