H: Dr. Banerjee, can you explain how AI is currently being used in the fashion industry, particularly in forecasting and trend analysis?
DR.B : There are three different pillars, within machine learning: [AI] supervised, unsupervised, and reinforcement learning. All these are being used in different ways, for example in forecasting and predictions, you have a data set.
Imagine a problem like weather forecasting, how do you predict the weather using data? By looking at certain combinations, whether it is sunny or rainy is a particular outcome which depends upon a certain combination of features. If features align in a particular way, it will be sunny, if the features align in a different way, it will be rain. Therefore, if you have a lot of data with you, there will be a different outcome if you train a model than if you do it with the human mind.
In the same way, the fashion industry has a resemblance with a similar forecasting model.
The different types of predictions are whether a product will sell or not, and it will depend upon a certain combination of features from the past. There are number of entities on an E-commerce site, and you have certain combination of independent variables and one dependent variable which you’re trying to predict.
You can divide the consumer tastes and preferences within certain clusters, it is basically a penalty and reward-based model, and you can train the model based on these two features and the goal of the model is to maximise the reward.
For everything there is a mathematical algorithm, therefore what we are doing is we are combining everything and looking at it from a mathematical standpoint. Even the human mind does the same thing unconsciously.
H: AI has made it easier for the fashion industry, looking at trend forecasting in particular?
DR.B : There are at least 50 companies in the world who are purely working around trend predictions. They are trying to understand what is hot in the market, and the easiest source of data is E commerce. It’s basically web crawler who ‘crawl’ the E-commerce pages of competitors worldwide or within a specific geography.
It’s free from any copyright, as the data is in public domain. They identify, for example on Amazon, with images, price points, comments etc. If you have information of hundreds of brands, AI can tell you how they’re changing and how their prices are reducing. Trend forecasting is nothing but looking at the past.
H: How do you see AI contributing to more sustainable practices in fashion, especially in reducing waste and improving efficiency?
DR.B : The fashion industry has a waste problem, as we create more waste than value.
The supreme reason is because brands don’t want to create waste, they only want to create profit. In the process of making the profits, the wastes are also created now using machine learning through sophisticated algorithms.
We are helping the brands become profitable by optimising and minimising their inventory.
By maximising their profits, increasing their full price sell through, identifying the right products, prices, time, places. Any type of analytics, whether it is descriptive, predictive, diagnostic or prescriptive, are helpful in terms of providing them with insights. And when you take decisions based on insights, they’re always better than any decision without.
So, I think overall, the background of analytics and AI is all about creating optimising and making things simple and easy for brands.
H: What are some of the biggest challenges fashion brands face when incorporating AI into their operations, and how can they address them?
DR.B: There are a large number of challenges at the present moment, notably being the lack of technical knowledge. Everyone wants to do it, but nobody knows how to do it.
Last year there was a report in The World Economic Forum, stating that with 75% of the brands who invested in AI, only 25% of them could actually realise the benefits of their investments.
So, I think the number one challenge is the technical as the amount of progress that we have been able to make is almost like 30 years of progress we have captured within three years’ time. Things are fast paced, so people are not ready for it, and business transformation is a major challenge.
The third reason could be the technical infrastructure, because most of the developments are with a few companies. So going forward, I think brands need to have their own technical infrastructure and data structure.
There could be concerns when it comes to wars, as the world is becoming more unpredictable and if the technology lies in few hands, there is a risk for the whole world. Hence, I always talk about democratisation of technology, where all the companies have smaller technical setups everywhere. The data being shared and scattered across the globe, so I think going forward, every country will need to have their own data centres.
So, in the future, if a war like situation happens, it does not become a data war.
Now, take a look at inspiring stories of Upcoming Fashion Brands !

