How to Implement Artificial Intelligence in Fashion Retail

Defined as a technology that has the ability of computers or machines to perform tasks that are typically human, Artificial Intelligence (AI) is a technology that is developing rapidly. As it continues to advance, now is the perfect time for fashion brands to adopt AI. Those who do not get on board now will probably realise the value of AI too late, and fashion brands using Artificial Intelligence would have already taken the technology to another level. In this article, I will be making several recommendations that can contribute to the successful implementation of artificial intelligence in the fashion retail business, and I will also be discussing the pitfalls.

Image Credits: Syte

I have asked Jelle Stienstra, Digital Strategy Director at PTTRNS.ai, about AI in the fashion industry today, he shared:”AI is an integral part of different industries these days, more and more companies are using the fast-developing applications of AI, and many first movers have embraced AI technology and are applying it for different purposes”.

While sophisticated, there is often much unfamiliarity with the implementation of Artificial Intelligence in fashion, which can make this retail tech project risky. Several steps must be taken before the AI application is operational, but organisations can prepare themselves for this. Both large and small companies in the retail and fashion industry have started implementing and applying various types of AI applications. A typical example of this is image recognition. Image recognition is a term for computer technologies that can recognise specific people, animals, objects or other targeted subjects through the use of algorithms and machine learning concepts. The image recognition application must be trained as well as possible to make connections between different images.

There is an essential factor that cannot be grasped with AI, and that is the haptic element: the “look and feel’’ of the consumer. About this, Stienstra says: “Fashion is about feeling, but AI is not ready yet to recognise feeling”. Gerard Smit, Benelux Chief Technology Officer & Digital Transformation Leader at IBM, agrees. He states that the use of sentiment and emotion analysis is not utilised to the maximum in the fashion industry. Smit mentioned: “Emotion analysis is about how people feel. There are already several chatbots that can analyse a feeling and emotion through the way the consumer communicates”. Smit knows that there is a significant development going on to improve the online feeling. When the consumer buys something online, he or she does not yet have the opportunity to feel whether the substance suits them. A new development of technology, which is not yet on the market, gives the possibility for the consumer to ‘feel’ the structure of the substance of rubbing the screen of a smartphone. My research among Dutch consumers has shown that they are looking for an ‘experience’. This new development gives consumers the experience they are looking for.

AI Fashion Retail Solutions Implementation Steps

The implementation of an AI system – for example, a service robot or a machine learning application – is not much different than the implantation of, for example, a standard IT business solution. Nevertheless, there are several necessary steps to take that makes the difference between a successful or a failed project.

Before you follow the steps, there are several necessary conditions that the implementation project must meet. The strategic perspective of AI is critical because the ultimate goal is to stimulate business value. Through the use of artificial intelligence and its application in the apparel industry, organisations put themselves in a position to better understand the customer and serve them more efficiently.
The management within organisations must not only be involved but must also set clear goals and expectations concerning the desired benefits of the AI project.

“Business leaders and project managers must first devote time to clearly defining and articulating specific problems or challenges that they would like to have solved by AI.”

There are various project management methods for implementing AI, such as the Waterfall method. The steps follow each other, and the process is followed from top to bottom. A negative about this method is that it is difficult or impossible to come back on steps that have been taken. In addition to the Waterfall method, there are more modern methods to implement AI such as Agile. However, Design Thinking is also very suitable for an AI project. Design Thinking is a method to realise new solutions and new ideas creatively. The use of innovative power in the organisation through brainstorming sessions, the development of multiple prototypes, the continuous improvement of prototypes and the deep thinking of the organisation, makes Design Thinking a unique methodology for developing and implementing AI.

In my implementation example, I applied the Design Thinking method, but Waterfall or other methods are also applicable. There are 7 steps that a retailer must follow to implement AI in the company.

7 Steps a Retailer Should Follow to Implement Artificial Intelligence in Fashion

The first step is to define a clear use case. This is an integral part in determining the success of AI. Business leaders and project managers must first devote time to clearly defining and articulating specific problems or challenges that they would like to have solved by AI. The more specific the goal, the higher the chance of success. Consider, for example, business management ratios: x percentage higher conversion when ordering on the website, x percentage fewer complaints, x percentage more turnover via the website, x number of fewer returns, x more turnover per customers, etc.

The second step that an organisation must take is to increase awareness within the management team. During this step, it is crucial that the organisation becomes familiar with AI. AI is an abstract concept for employees, and they are not familiar with the possibilities and impossibilities of this relatively new technology. Therefore, take the time to train the organisation and become familiar with all facets of AI.
There are sufficient training courses to introduce the organisation to the various applications of AI. An example of training is the basic speech recognition training from the provider IT4Speech.

Also Read: Why FashionTech Fails? Part I: Legacy

The next step that follows is developing a hypothesis. Extensive tests must be performed to determine which variables or characteristics are most significant for an AI application or algorithm. This comprehensive test validates the hypothesis and assists to validate the final choice and to improve the implementation of a prototype algorithm or AI system.

A critical step is step 4: required & available data. Existing processes and systems must be able to access the data and record the data needed to perform an analysis. AI works best when large amounts of ‘rich data’ are available. It must also investigate whether the available data tells the true story; this is called data validation. Large amounts of rich data give AI project a more considerable change of fine-tuning and the quality of predictions is therefore rapidly increasing. The data quality is all-determining for the success of an AI project. If you start with bad data, your AI project won’t be a success.

Step 5 is to develop a prototype; for this purpose, a hypothesis validation methodology must be developed. This methodology defines the performance measurements, test data and test protocols. The validation methodology is also crucial when comparing, analysing and testing the various prototype algorithms. Later in production, this validation technology can help to refine and further develop the algorithm. A particular recommendation is to start with small amounts of data to test, validate and improve the prototype.

An AI or machine learning system requires a modern architecture computing power, storage capacity and speed.

During the next step, it is vital to start building and testing. It is advisable to develop different types of interfaces for different types of users, such as business analysts and end users. For end users, dashboards are the appropriate way of communication with the AI system. For business analysts who need to monitor and assess a machine learning model or robot, other interfaces could be developed. An AI or machine learning system requires a modern architecture, computing power, storage capacity and speed to be able to process the vast amounts of data. Many processes and systems do not have these new features, and therefore, many organisations turn to the Cloud. There are could suppliers that can deliver AI as a service from the cloud: such as Amazon, Google, IBM, Microsoft etc.

Step 7 is about the aftercare of implementing AI. During this step, it is easy to assume that the project has been completed, but then the so-called ‘learning time’ starts. The organisation must permanently monitor the algorithm for the correct functioning and must continuously monitor whether it functions as desired. Market conditions and business models change at lightning speed, and the algorithm needs to be adjusted accordingly.

A Few Recommendations

An important recommendation is getting the employees involved when applying AI. The role of employees in the workplace is changing, and employees need education in applying AI. AI projects require specific technical skills and many organisations do not have this available in the workforce. This is not just another standard automation project, but a complex project with many different and new disciplines: think of data science, integration specialist, consumer behaviour specialist, etc.
It is, therefore, essential to purchase or hire this knowledge. There is much demand in the market for these new competencies, and the consequence of this scarcity is high personnel costs.

“A critical recommendation is linking consumer data with the product range.”

Another important recommendation is that your data needs to be complete, correct and relevant. A good example of incorrect data are the different forms of writing of a simple word such as vanilla. The consumer can write the word vanilla in endless different ways, and for AI, it can be challenging to recognise the different interpretation of the word vanilla. As a result, the application is not functioning correctly.

A critical recommendation is linking consumer data with the product range. This link gives a lot of customer insight. The program is much more accurate than humans and can process much more data in a short time.

Responding to different types of languages use, for example, street language, is a recommendation which offers the possibility for greater recognizability among consumers. Depending on your target group, it is advisable to provide various options to allow the consumer to respond in the language they use. Greater recognizability among consumers leads to higher conversion.

Loading...

Loading…

Today’s consumers are critical and are looking for transparency. An organisation can respond to this through ethical codes of conduct for AI and data privacy. An example of transparency from an organisation is the naming of the use of personal data.

Consumers select items by feeling, and it is difficult to find out what consumers like most. So many fashion retailers are found online with many different products on offer, making it more difficult for consumers to find something suitable from all these different items. AI is then there to make recommendations, for example focused on size, body shape and house colour undertone. This is to understand the ‘feeling’ of the consumers. AI is not yet fully developed and still has difficulty recognising this feel.

Also Read: Facebook and Farfetch Join Forces To Launch Libra, a Cryptocurrency Built on the Blockchain

Last but not least, the most important recommendation is that there is no reason to wait with applying AI. If you do not use AI within an organisation, your competitors certainly will. AI is here, and if you do not choose to use it, you might be too late.

Share Your Tips & Corrections

Mano ten Napel

Founding Managing Editor at FashNerd.com
FashionTech Consultant, Freethinking Opportunist Hero, Aspiring to Inspire. Wearobot Groupie with a Tech for Passion. Contributed to Wired.com. Founder of WearableGuru.com Founding Managing Editor FashNerd.com
Mano ten Napel