Greg Harris for Net-a-Porter

How artificial intelligence is powering retail customer experience

Retailers are beginning to explore how cognitive computing and AI could make e-commerce smarter and more personalised

According to analyst Forrester, artificial intelligence (AI), big data and analytics will increase businesses’ access to data, broaden the types of data that can be analysed, and raise the level of sophistication of the resulting insight. For 2017, Forrester expects investment in AI to triple.

“AI will provide business users with access to powerful insights before they are available to them, through the use of cognitive interfaces in complex systems, advanced analytics and machine learning technology,” says Forrester principal analyst James McCormick in the company’s Predictions 2017: artificial intelligence will drive the insights revolution report.

Forrester expects that AI will drive faster business decisions in marketing, e-ommerce, product management and other areas of the business by helping close the gap from insights to action

Customers increasingly want retailers to offer convenient, responsive and personalised services. A recent IBM study revealed 48% of customers believe it is important for retailers to provide on-demand personalised promotions when online, while 45% want the same options in store.

Customers also want to discover and purchase products however, whenever and wherever they want and are assuming ever-greater ownership over their retail journeys.

Today’s customers have almost limitless choices in their discovery and inspiration. They are not constrained by time of day or location when making purchases, IBM’s study reported.

The human brain can consume and process only a limited amount of information while traditional computing is pre-programmed and rigid, unable to learn, reason, relate or interact in natural language, according to IBM.

In IBM’s survey, 91% of retail executives said they believe cognitive computing will play a disruptive role in the industry, and 83% believe it will have a critical impact on the future of their businesses.

In 2016, outdoor clothing retailer North Face began using Watson, IBM’s cognitive computing platform to power a personalised shopping experience for customers on its website.

Read more about cognitive computing

  • Ahead of its annual meeting in Davos, the World Economic Forum warns that artificial intelligence needs strong governance.
  • IBM is promoting its Watson natural language processing analytics technology as it tries to move its core business beyond technologies that it pioneered, but which have become commoditised. Will it fly?

The retailer used Expert Personal Shopper (XPS), a shopping platform from Fluid, which IBM acquired in November 2016.

XPS runs Watson’s natural language processing to help consumers discover and refine product selections based on their responses to a series of questions.

For example, after a shopper enters details on a desired jacket or outdoor activity, XPS will ask questions about factors such as location, temperature or gender to recommend a suitable jacket.

Another company looking at exploring the possibilities of cognitive computing and AI in retail is luxury fashion retailer Yoox Net-a-Porter Group (YNAP).

Teaching machines to solve problems

Roger Mitchell, product and platform director at YNAP, says: “AI will undoubtedly change our business.”

AI is a change in mindset from traditional programming. If programming is about telling the machine what to do through the instructions in the programming language, then the machine must be taught something to get the desired results, says Mitchell.

If, for the sake of argument, traditional programming can provide a 1% improvement, Mitchell believes AI could solve the same problem with a possible overall improvement of 10%.

The reason Mitchell believes AI will offer a clear advantage over traditional programming techniques is that it offers a way to tackle problems that would be hard to solve any other way. Cognitive computing is evolving quickly, he says.  

“There are elements of AI that represent a leap in faith, but others elements are proven. We know some of the APIs are pretty mature and have been proven for customers in financial institutes, where they can improve customer care by reading correspondence or they can understand tone of voice to evaluate customer sentiment,” he says, but adds that the technology still needs to be adapted for retail.

“There are some other areas, the alchemy areas of Watson such as the automatic storage, sorting and classification of any type of image or any kind of document, which is probably a little more experimental, but it’s a huge opportunity,” he says.

What this means for a retailer such as Net-a-Porter is that there is a continuous stream of cognitive activities that it is able to apply immediately given the technologies it has access to, as well as others that the retailer will grow into as its e-commerce platform is opened up to AI.

For Mitchell, AI represents a huge opportunity for Net-a-Porter, and he believes the IT team is in a strong position to take advantage of it.

Given cognitive retail relies on teaching the machine, he says: “You need to have a clear idea of your business problem and you need the time, patience and investment to get the results.”

At the end of January 2017, YNAP ran a two-day hackathon, supported by IBM, to explore the application of AI in retail.

Among the ideas was one where the shopper can upload a picture of a celebrity and the system then finds the celebrity’s outfit or a similar one to purchase.

Another idea tackles the issue of returns by understanding what clothes will fit, based on a deep understanding of the size of garment from fashion houses.

Others explored how to reduce the number of returned items, and a number of teams tried out natural language and voice-based user interfaces.

Using AI to get closer to the customer

The hackathon was the first retail-themed one to be held in Europe.

Chris Williams, IBM Watson, chief architect for Europe, says: “Cognitive fits into retail primarily in the area of providing better customer journeys. You can use it to make recommendations and guide people through the buying process by having more expertise embedded into the web or app experience.

“There are also opportunities to use it in the back office, such as handling images and other content, which is pretty important in retail, particularly around classifying products and understanding what appeals to people.”

While retailers can amass a vast amount of structured data about their customers through their web browsing habits, what they search for and the kind of products that interest them, Williams says: “Cognitive computing is key to unlocking the understanding of unstructured data.”

An example could be understanding the meaning of text or identifying an image. “What is the image showing and what are its attributes – which, in a retail environment, ties in with the attributes of the product itself, from simple things such as colour, style and other attributes,” he says.

“The more understanding you have in the system, the better it can be at making recommendations and guiding customers.”

This begs the question of where traditional business intelligence and analytics gleaned from transactional systems fits in. Williams says: “A typical pattern is combining insights from Watson with the more traditional data mining and predictive analytics.”

In this respect, he says Watson provides more data points to use in traditional data models.

The challenge today is that businesses are only now starting to understand how to use big data. cognitive computing represents an altogether different set of information sources that data scientists must grapple with and make sense of.

For Williams, this starts with a discovery process to help data scientists uncover data sources they may otherwise have missed. “The skills of data science do not go away in the cognitive world. It is about bringing together what’s relevant.”

He believes it is essential that business and technical people collaborate closely to solve real business problems and to have success with cognitive computitng.

“It is possible to start small, since the Watson platform is a set of services that you can combine together in many ways. What we will often be looking for is in parts of the business where there’s a high reliance on unstructured data and some level of expertise is needed to interpret that data,” he says.

“If we can automate some of this interpretation then we can make the business process more efficient and help people make better decisions. This is where we see the quick wins in cognitive computing.”

AI is the next step in IT

IBM has been commercialising Watson for more than three years and AI represents one of the biggest growth areas for the company, as it evolves its core business.

Healthcare has been the number one application area for Watson but, as YNAP and North Face have shown, cognitive computing could offer retailers a new way to improve customer service.

In Williams’s experience, It is possible to have a proof of concept cognitive system up and running in a matter of weeks. But, to get the best from a cognitive system, it is often necessary to bring together multiple source of information, which can take more time.

“In the prototype, we often simulate the integration with a view of plugging in the real data sources further down the line,” he says.

Projects often start by providing the AI with deep domain expertise in one specific area, which can then be expanded out. The other aspect of a cognitive project is how people will interact with the system, the questions they want answered and the type of language they will use.

Clearly, there is a spectrum of complexity ranging from cognitive projects that can be achieved today to those that are likely to remain science fiction for the foreseeable future.

For any organisation exploring the possibilities of AI, experts stress how important it is that business stakeholders grasp the breadth and depth of the problem the cognitive system needs to solve and the amount of data and content that needs to be fed into the system to achieve this objective.

Next Steps

AI-powered retail systems create opportunities for channel partners

Read more on Artificial intelligence, automation and robotics