When the coronavirus pandemic struck, closing down economies and travel in a way no-one had thought possible, humans had to adapt fast.
But it was not just people who needed to learn a new way of life, with the artificial intelligence systems upon which so much industry survives also requiring a little helping hand to come to terms with the dramatic shifts.
It has emerged that many AI systems which control inventory, marketing, retail and even fraud needed tweaking because they could not compute why the lives on which they made their predictions had changed so much.
Fraud detection services were among those that had to be told that a surge in purchase of power tools were simply people doing lockdown DIY.
And as top searches on Amazon switched from Lego and phone chargers to hand sanitizer and face masks, the company appeared to switch the algorithm so products that they ship were not always the top recommendation.
David Excell, founder of Featurespace which uses AI to detect fraud for banks and retailers, said that social distancing measures across the world has seen a reduction in a lot of activity as people’s spending footprint changes.
Instead of spending money in bars and restaurants people are buying groceries and shopping in off licences. Instead of foreign holidays people are buying home improvement items.
Mr Excell said that whilst their systems have adapted well “there has across the industry been the need for some manual intervention.”
One tweak which they needed to make was for the increase in spending on DIY online, he said, adding: “Power tools are ideal for fraudsters because they retain a lot of their value and have a high resale price.”
Dr Michael Feindt, Strategic Advisor to Blue Yonder which uses AI to provide supply chain management and retail planning, said that as soon as lockdown was imposed there were “big changes” in customer behaviour which have continued to change from one week to the next.
He said that he was pleasantly surprised with how well their systems adapted, but they had discussions with customers on how they “turn the steering wheel” and they made alterations to account for a complete shift in purchasing.
The pandemic has proven that the algorithms need not only to be “robust” but “able to change” quickly, Dr Feindt said, particularly as different parts of the world react to events in different ways.
After the crisis began people in the US started looking to buy guns, the French red wine and Britons searched for toilet roll and hand sanitizer, he pointed out.
But the industry will also need to be aware of the changes that will happen as countries gradually emerge from lockdown, it is said.
“Fraud is an interesting challenge of social distancing, as it is relaxed and businesses start to reopen and people start going to sporting events and start travelling again you need to make sure you are able to pick back up to normal spending again,” Mr Excell said.
Even when the world returns to a new normal, habits will have changed, with research by Blue Yonder finding that two-thirds (64 per cent) of shoppers who are spending more online say they will continue to do so once the crisis subsides.
Since the pandemic began stock availability has overtaken price and range as the most important thing for consumers.
And just as humans will have to adapt to a world post-coronavirus, machines will also have to take into account the fact that 2020 has been an anomaly in making their predictions for the future.