A google search for ‘AI’ returns 670 million results. In comparison, a search for ‘retail’ yields only 380 million results.
The hype surrounding artificial intelligence is undeniable. It’s making headlines, cropping up in conversations, and being discussed in boardrooms to decide if, when and how companies should be making use of this new technology. But what does this mean for retailers? Is artificial intelligence just ‘hype’ – short-lived, over-discussed and under-utilised – or will it really transform business models and change the way we live and work?
How AI is being used in retail
The potential for AI and its wider set of related technologies to increase efficiency and improve the customer experience is well known. The last two posts in this blog series demonstrated some of the ways that AI can be used across the entire consumer journey; from chat-bots learning about your style and size to make personalised product recommendations, automatic stock replenishment ensuring the products you want are available, right through to driverless cars and drones delivering the products you choose to buy. And as the technology advances, the uses for AI are growing beyond the conventional and into the sci-fi domain.
Amazon Go in Seattle uses cameras and machine learning algorithms to scan which items consumers pick up during their shop so that consumers can walk out the store with no checkout process. KFC is using AI facial recognition software to provide menu recommendations based on variables like gender and facial expressions.
Challenges in implementing AI
Most retailers are aware of AI and its immense potential to help them offer a better customer experience. The reality of integrating it with their day-to-day operations is, however, like any new technology, more complicated. According to a recent report by MIT’s Sloan Management Review and BCG, 85% of surveyed executives said they believed AI would give them a competitive advantage. But the actual adoption of AI doesn’t match up. Only 20% of surveyed companies used AI in some way, and only 5% employed it ‘extensively’. For the hype surrounding AI to become a grounded reality, companies must learn how best to use the technology and deal with the inevitable complications along the way.
Building the foundation & harnessing data
The first hurdle for retailers is data. Having an effective AI strategy relies on gathering and structuring vast amounts of data, a time-consuming task which is further complicated by data ownership concerns. Andy Done, director of data engineering at Farfetch, warned that “no impressive algorithms or technologies can overcome a lack of high quality data” and that “retailers cannot expect to drop AI on top of a big data mess”.
Given how quickly AI has gone from vision to reality, there is a shortage of people who can understand and use the data. Whilst this problem may in part be addressed by the likes of Google’s AutoML (a machine learning software that can itself design machine learning software), the increasing proportion of the workforce using AI in their jobs must be able to understand and trust the decisions made by this software. This will require not only practical support and training, but also a mindset change. Workers will need to let go of their old ways of working and embrace change. With many fearful that AI will replace their jobs, implementing a strategy with humans and computers working in harmony will be no easy task.
Experimentation & partnerships in a ‘sandbox’ environment
The second challenge for retailers is to create an effective tech environment where they can test, learn and innovate. Whilst the cost of error in retail is potentially lower than in healthcare or financial services, large amounts of time and money can be wasted if initiatives aren’t tested and iterated before being rolled out. Financial services firms have successfully created ‘sandbox’ environments to securely road-test artificial intelligence products and services, and forged partnerships with disruptive market entrants. This has allowed fintech- and AI-focused startups to quickly deploy and test solutions on a small scale.
Retailers must learn from the financial services sector to create an environment similarly open to innovation and partnerships. Sia Houchangnia, investment partner at Seedcamp, warns however that “for these partnerships to be successful, startups must demonstrate that they can help large corporations tackle specific business needs, rather than offering general ‘AI solutions’”.
Making choices of when & where to use AI
A final challenge for retailers is the question of where, when and why AI should be used. Retailers might be tempted to use AI anywhere and everywhere, to enhance every possible process. Prioritisation of opportunities will be key to identifying which experiments to run. A good litmus test is typically value for the customer, whether that be streamlining their experience or personalising it.
Danielle Haugedal-Wilson, head of architecture and analysis at Co-op Digital, warns about the dangers of using the technology for its own sake. “The focus should always be on making the experience better for the consumer.” For example, while personalised product recommendations and user journeys may benefit the consumer most of the time, a balance is needed. If suggestions become too narrowly personalised based on previous behaviour, people may miss out on brand new products and experiences if they don’t fit their (machine learnt) preferences.
Another good criteria is increasing efficiency and accuracy in merchandising and supply chain decisions, such as pricing, promotions, forecasting and replenishment – tasks that often rely on macros on a spreadsheet. Algorithms can now get the job done much quicker, more accurately, and consistently. Again, this needs to be balanced with value for the customer. Offering too many price promotions and coupons can become confusing, annoying and potentially even creepy, despite the potential cost saving benefits.
Now is the time to act
Whichever of these applications retailers choose to do, the point is, they need to be experimenting with them now.
AI is not changing the retail landscape – it has already changed it. Paul Clarke, Chief Technology Officer at Ocado, spoke of AI as a ‘technological tide sweeping over the world which businesses must respond to or be washed away…’ “It’s going to come a hell of a lot faster than people think; it’s overhyped and underestimated at the same time.”
He points out, “What always mystifies me is when people debate about whether we should embrace new technology – it’s a pointless question. There is no choice… if we don’t embrace it, others will… It’s about getting with it and fast, rather than holding back and pondering.”
But for retailers to derive real value from AI, they must address the challenges already prevalent in adopting an AI strategy, as well as preparing for as-yet unforeseen challenges that will inevitably arise as the technology develops. And if the first mover advantage is to be anything like that enjoyed during the rise of the internet in the 90s, retailers need to act fast and tackle those challenges head-on.