The rise of checkout-free shops and chatbot estate agents


Amazon has long employed machine-learning, a type of artificial intelligence that recognises patterns, to serve shoppers’ preferences.

But other online retailers are now investing in this personalisation. Shop Direct, which operates the and Littlewoods brands, says AI is its “big bet”, enabling the personal relevance needed to capture flighty mobile shoppers.

“We now have just three seconds to grab a customer’s attention on a five-inch screen,” says Sasha Berson, customer and strategy director, explaining that AI is essential to personal relevance. “If we don’t serve up thumb-stopping content, they’ll swipe on.”

Evermore powerful processing also lets machine-learning decide timely e-mails for hundreds of thousands of individual customers. Mr Berson explains: “Our retention engine listens for the signs of ‘lapsitis’ – behavioural patterns that indicate falling customer engagement – and automatically contacts customers with a relevant incentive to shop again.”

Last year, Shop Direct launched an in-app chatbot called the Very Assistant, in which users can talk to a virtual customer agent through their phone. It plans to turn this into a full personal shopping assistant that can learn each customer’s needs and advise on purchases.

Meanwhile, online takeaway business Just Eat uses AI to personalise recommendations and to enable customers to make voice orders through the Amazon Echo home AI device. Adrian Blair, Just Eat chief operating officer, is reported as saying AI shows “the way to the future, making things unbelievably easier for the customer and more efficient for the provider”.

AI is far from exclusive to online businesses. Starbucks is testing a chatbot in its app that enables customers to place an order conversationally, then collect it at a café. Japanese fashion retailer Uniqlo and US department store chain Macy’s are preparing in-app AI to advise customers browsing their shops on styles.

There is one bricks-and-mortar type of AI likely to dominate called pick up and walk out. Amazon’s first food shop, in Seattle, uses sophisticated machine-learning to enable customers to walk in with their phone, then simply pick up items and be charged. The technology identifies what is taken and by whom.

The company says on its website: “Our checkout-free shopping experience is made possible by the same types of technologies used in self-driving cars: computer vision, sensor fusion and deep-learning.”


Ai in oil and gas

In recent years, the oil and gas industry has looked to AI to improve customer experience, with chatbots at petrol pumps and online.

But the big change will be behind the scenes, in exploration and drilling. Management firm McKinsey expects $50 billion of savings will be made across the industry by using AI to improve exploration, well development and other processes.

Shell has installed sensors in thousands of oil wells for better understanding of their status. This is then analysed by AI and presented in 4D maps to the crew. According to a spokesperson, the volume of data “is of course impossible to understand correctly if not properly visualised” in this way. The company is also sponsoring a $7-million prize for technologists creating devices that can search smartly for oil below 4,000 metres deep.

Elsewhere, oil field giant Halliburton says its smart workflow automation has enabled operators to  increase production by 7 per cent through learning what is working and informing human decision-making.

BP, which collects extensive information from rigs and other sites, expects AI to be crucial. “At one level, AI offers a significant opportunity to maximise the use of the vast amount of data we collect on our operations, to optimise the many physical and commercial processes that are critical,” says Morag Watson, vice president for digital innovation.

AI can handle “complex situations such as computer automation of the analysis of pipeline video inspections”, she says. Other applications include interpreting reports created by operators and engineers.

Ms Watson expects cognitive computing, an advanced form of AI, to augment employees’ decision-making by absorbing their knowledge, “allowing computers to learn and behave with human-like reasoning in the many technical domains, while assimilating more data than a human can possibly hope to”. Computers could become the constantly available technical experts “making exploration more successful, well designs more effective, and drilling faster and safer”, she says.

The challenge is achieving cultural fit. “The advent of big data and analytics allowed us to take a different approach to interpreting our large volumes of data, and this takes time to permeate an engineering culture,” says Paul Stone, senior technology principal at BP. “AI and cognitive computing are showing us we can, and need to, bring engineering and data science principles to work together.”


AI in real estate


The property and land sector sees huge potential for AI in simplifying bureaucratic procedures. Buyers and their property managers are working with law firm Mishcon de Reya’s machine-learning to improve transactions.

“Historically, people would think about sales as a flood of documents, agreements and leases,” explains Nick West, chief strategy officer at the firm. “But ultimately the process is about data and we can automate a huge amount.”

The technology enables customers to understand a deal quickly. A firm considering buying a property and being faced with huge documents can count on machine-learning to read and interpret contracts, and help determine the quality of investment. “It can also identify any problems in the lease, highlight rules around rents and find licences needed for alterations,” Mr West adds.

The company is so keen to make the most of AI that its in-house technology incubator, created to identify startups, has shortlisted six machine-learning companies among twenty of interest.

Other firms are using AI to help find property. CityBldr a startup, rapidly locates good under-used sites for developers. Its system scans properties and ranks them, an entire process that typically would have taken experts more than a month.

The system accesses 118 million data points for buyers, including other sales, permit applications, planning data, land size and topography, traffic and location. Equally, it helps sellers see whether commercial developers or house builders would pay more.

Bryan Copley, CityBldr chief executive, explains: “AI has allowed us to create a marketplace specifically for under-used real estate. Humans’ insights are imperfect, but a machine can get better all the time. It draws in over 200,000 updates overnight.” The system is currently available in Seattle but is expanding to Los Angeles and then to a raft of other American cities.

Chatbots are also increasingly important in eliminating lengthy processes. Startup Apartment Ocean has created bots that enable estate agents to answer inquiries and shape good leads. “Real estate firms use the system to secure up to five times as many leads, asking the questions they want,” says president Nick Kljaic. “The system automatically captures the data, converting it into a lead even when the agent isn’t there.

“People love to chat. Now they can have a personal conversation at any time.”