The AI chatbots like ChatGPT amongst many others have gained rapid popularity in no time, sparking a race to develop similar competitors. Unlike conventional data categorisation, generative AI generates novel information using foundation models like GPT-3.5 and DALL-E, which excel at complex tasks simultaneously. Despite the fashion industry’s foray into AI, metaverse, NFTs and AR/VR, it has limited exposure to generative AI as of yet.
Although this new tech has teething problems, its potential to transform various aspects of fashion business is evident. Forecasts suggest generative AI could contribute US $ 150-275 billion to fashion sectors in 3-5 years. According to McKinsey, fashion brands and companies invested approximately 1.7 per cent of their income in 2021-22 on emerging technologies like generative AI and this is estimated to rise to 3.0 per cent – 3.5 per cent by 2030.
What actually is generative AI in the fashion industry’s context?
Generative AI constitutes a diverse array of tools and software that collectively serve the purpose of identifying items that closely resemble or subtly deviate from an established dataset. Within the context of the fashion industry, this technology addresses the persistent challenges faced by retailers in terms of varying consumer preferences and purchase decisions. Within the realm of fashion retail, this challenge translates into presenting customers with a plethora of choices – be it an attire item available in a specific shade of blue or an alternative variant featuring elongated sleeves, among other considerations or requirements. So how does a retailer harness the generative AI tool to give shoppers that personal touch is the main catch here.
Not just personalisation, generative AI can help fashion brands and retailers in a lot of areas such as merchandising, sampling, product development, trend forecasting, supply chain, logistics, store operations, consumer experience and even marketing.
How do quicker sampling and product development work via Generative AI?
The conventional cycle of product development involves designing, sampling and manufacturing, which usually spans 30 to 180 days. This cycle has proven insufficient in today’s rapidly changing market dynamics. This has led to issues like excess inventory, shrinking profit margins and a dissatisfied customer base.
Now imagine a timeline, compressed to a mere 90-96 hours, that encompasses catching a trend as it rises, designing and manufacturing at lightning speed. Many SaaS platforms are slowly catching up with the market where they connect different parts of the fashion industry like brands, designers and suppliers.
Some of them have more than 30 to 50 suppliers ready. This dual-powered network caters to both (ODMs) who produce their own designs and (OEMs) who develop designs based on the tech provider’s specifications. The Generative AI-powered design platform provides a bucket of trend insights and a vast library of product visuals, empowering suppliers to make decisions on which design to produce.
One of the examples is Zapero – a Hyderabad-based fashion tech SaaS provider – that helps make design quickly with its vast network of agile factories that cater to small batches. Notably, the company focuses on two important parts ‘Generative AI Design and Virtual Samples’. Due to this, it helps brands react quickly to new trends and the preferences of people who like to shop online instead of following a slow calendar and guess work.
Furthermore, with tech providers’ agile approach to design, production and replenishment, brands seize opportunities as they cater to the Gen Z-driven market where new styles are required every week resulting in new designs every month instead of coming up biannually which was the scenario earlier.
KC Puttagunta, Co-founder of Zapero, in conversation with Team Apparel Resources (AR), asserts, “By harnessing the prowess of Generative AI, integrated commerce and a robust supplier network, Zapero propels fashion brands into the era of rapid response, profit optimisation and customer delight. It’s not rocket science; it’s the magic of innovation aligned with the art of fashion.”
| Zapero propels fashion brands into the era of rapid response, profit optimisation and customer delight. It’s not rocket science; it’s the magic of innovation aligned with the art of fashion. KC Puttagunta Co-founder of Zapero |
Generative AI algorithms can predict fashion trends and assist in demand forecasting
The fashion trend predictions use a variety of statistical models. In layman’s terms, generative AI algorithms look at all the past historical trends and ‘near-to-accurately guess’ what is coming up next. As the model gets proficient, it starts generating new fashion designs, descriptions and styles that align with the trends it has learned from the training data. For example, it might generate descriptions of clothing items with unique combinations of colours, materials and embellishments that are emerging as popular trends.
Instead of relying on trend reports and market analysis alone to inform designs for next season’s collection, both mass-market fashion retailers and luxury brands’ creative directors can use generative AI to analyse in real-time various types of unstructured data. Generative AI can, for example, quickly aggregate and perform sentiment analysis from videos on social media or model trends from multiple sources of consumer data.
So, how does this whole process work? Creative directors and their teams could input sketches and desired details – such as fabrics, colour palettes and patterns – into a platform powered by generative AI that automatically creates an array of designs, thus allowing designers to play with an enormous variety of styles and looks. A team might then design new items based on these outputs, putting a fashion house’s signature touch on each of the looks. This opens the door to creating innovative, limited-edition product drops that may also be collaborations between two brands.
“If you see many more blue T-shirts being purchased, then it is likely that blue T-shirts will be demanded. The newest generation of models can go much deeper now and you can guess the colours, patterns or even styles that define a trend,” explained Charles Allard Jr., Founder, Delvify – a prominent young fashion technology company, during a conversation with Team AR.
| Delvify is about optimising materials sourcing by using computer vision and natural language processing, so less iterative design and more traditional AI. Charles Allard Jr. Founder, Delvify |
This is where generative AI facilitates the process of rapid prototyping in fashion design that follows successful trend forecasting. Most designers want to create new ideas and have them loved by consumers. The design is only one element, it is also about the choice – what do you choose and why? How does it fit in with the brand and its values! Rapid iteration may lead to new insight – all thanks to generative AI algorithms. “Delvify is about optimising material sourcing by using computer vision and natural language processing, so less iterative design and more traditional AI,” mentioned Charles.
Consumer experience can be boosted with generative AI
Fashion retail business is mostly about consumer engagement with brands and retailers’ offerings. And any tool that prompts more engagement such as a tool for computer vision to match similar items or a tool to allow customer to speak what they want and see some fantastical creations are all versions of engagement. Vans – a leading skate shoes and apparel brand based out of USA – does this quite well by engaging customers in store with creative designs for shoes. In future, they have plans to use a version of GAI (Generative AI) to show customers possible designs and translate those designs to shoes and related apparel.
Generative AI could also be applied to personalised customer communications. Companies that excel at personalisation increase revenues by 40 per cent compared to companies that don’t leverage personalisation, according to McKinsey research.
How are India’s fashion retailers positioned for using Generative AI?
Once a market that was considered a laggard in terms of technology adoption, India has emerged as one of the fastest growing fashion retail markets now – both in terms of revenues as well as technology integration.
The fashion-focused e-commerce platform Myntra has seamlessly integrated a ChatGPT-powered functionality within its shopping application. This innovative feature adeptly assists users in articulating their shopping preferences through natural and uncomplicated conversations. By doing so, this generative AI tool significantly streamlines the process of product discovery, sparing users the need to perform numerous individual searches.
| Some software and concepts, such as idea of virtual try-ons, have not worked very well in the fashion industry. Partly because it is slow and difficult to get right, partly because consumers like to touch and see garments in person. We may see companies like Visense or Syte branch out to generative content someday. |
Consider a scenario where users intend to make purchases for an upcoming wedding event. In such cases, consumers of Myntra can utilise Myntra’s advanced MyFashionGPT with prompts detailing their specific requisites. Subsequently, the tool rolls out a curated selection of outfit choices that align with the provided information.
Raghu Krishnananda, Chief Product and Technology Officer, Myntra, commented, “We believe this feature will redefine how customers will explore and embrace fashion, making every wardrobe choice a statement of individuality.”
The endless opportunities of generative AI are so lucrative that Zivame – a leading women’s intimatewear brand – has also opted for it. Zivame has done a successful pilot for venturing into intimatewear designs led by machines, and the brand believes these designs, coupled with the GTM efficiency of Zivame’s creation and fulfilment network, will bring a fresh perspective into what designing and launching lingerie products mean.
According to Monish Kaul, CTO, Zivame, conventionally, curation and selection are the keys to any platform’s offerings, to which the majority of brand promoters relate to. Machine-led intervention in understanding trends, curating high impact designs and generating a seasonally relevant catalogue will enable Zivame to create top sellers at a faster pace across its house of brands portfolio.
| Zivame has a data-driven belief system and most certainly generative algorithms are a core part of technology excellence given the possibility of applications across content, conversations & fashion innovation. Monish Kaul CTO, Zivame |
“Currently, we are starting with GANs and Auto-Encoder based ensembles to minimally launch and experiment with designs with a much shorter GTM lead time. Zivame has a data- driven belief system and most certainly generative algorithms are a core part of technology excellence given the possibility of applications across content, conversations and core fashion innovation,” commented Monish.
Future prospects!
While still nascent, generative AI has the potential to help fashion businesses become more productive, get to market faster and serve customers better. In future, it’s entirely possible that the designs showcased in Fashion Shows will blend the prowess of a creative director with the power of generative AI, helping to bring clothes and accessories to market faster, selling them more efficiently and improving the customer experience. All this depends on how brands and retailers want to make this technology scalable!
In the words of Monish, “There are many factors that are critical to the scaling process such as Sheer parameter and model size: more data implies more complex learning, which implies more parameters to learn and store in the model; Training generative models: these models perform better when quality and quantity of data points are abundant; and Commoditisation: Like all tech, much of what seems undoable today in terms of scale will soon come to life as a commodity service that companies can subscribe to such as training, provisioning and inferencing of generative AI models.”
