How to use AI to manage inventory in a time of stock shortages

A win­ter of retail dis­con­tent is loom­ing for con­sumers in the UK. Post-Brex­it bor­der delays and HGV dri­ver short­ages are already lead­ing to emp­ty shelves in super­mar­kets around the coun­try. There could even be no turkey for din­ner this Christ­mas. 

The Food and Drink Fed­er­a­tion has warned that shop­pers can expect not to find all the goods they want when­ev­er they want them for the fore­see­able future. Its CEO, Ian Wright, told an Insti­tute for Gov­ern­ment event in Sep­tem­ber: “The just-in-time sys­tem is no longer work­ing. I don’t think it’ll work again.”

Retail­ers and their ful­fil­ment part­ners are seek­ing ways to nav­i­gate this sticky sit­u­a­tion. AI could offer them at least some of the answers.

When peo­ple think about AI appli­ca­tions in retail, they might imme­di­ate­ly envis­age robots pick­ing and pack­ing prod­ucts at speed in a ful­fil­ment cen­tre, with drones hov­er­ing above them on the look­out for poten­tial prob­lems. These devices will indeed be a com­mon fea­ture in the ware­house of the future. But the AI-based tech­nolo­gies of most ben­e­fit right now are those that can offer busi­ness­es a bet­ter han­dle on their inven­to­ries and sup­ply lines. 

As shop­pers become more accus­tomed to see­ing out-of-stock notices, both online and offline, their pur­chas­ing deci­sions are like­ly to change – as will demand sig­nals. So says James Hyde, founder and CEO of James and James Ful­fil­ment, which han­dles ware­hous­ing and order ful­fil­ment for hun­dreds of small etail brands. 

Giv­en the ongo­ing delays to inbound stock, Hyde notes that there is “a great oppor­tu­ni­ty to revis­it the slow-mov­ing lines that are clog­ging up expen­sive ware­house space”.

When inte­grat­ed with enter­prise resource plan­ning soft­ware and inter­net-of-things tech­nol­o­gy, AI sys­tems can con­stant­ly mon­i­tor stock lev­els and adjust their demand fore­casts as stock flows in and out of a ware­house. 

Although ful­fil­ment com­pa­nies can’t con­trol delays upstream of them, they must do their utmost to min­imise any dis­rup­tion dur­ing a product’s time in the ware­house that might impede its progress to the cus­tomer, says Daniel Hulme, chief AI offi­cer at mar­ket­ing giant WPP.

“The inef­fec­tive­ness of ful­fil­ment is a pri­ma­ry cause of cus­tomer dis­sat­is­fac­tion. AI can help firms to achieve expo­nen­tial effi­cien­cies in ware­hous­ing, dis­tri­b­u­tion cen­tres and the last mile,” says Hulme, who sold Satalia, the AI con­sul­tan­cy he found­ed in 2007, to WPP in August. It has pre­vi­ous­ly worked with Tesco to opti­mise its last-mile deliv­ery process­es.

He sug­gests that com­pa­nies could use dig­i­tal-twin tech­nol­o­gy to improve their con­trol. In essence, a dig­i­tal twin is a vir­tu­al mod­el of a phys­i­cal asset – in this case, a ware­house – that updates itself in real time. “This liv­ing sim­u­la­tion can con­nect and orches­trate com­po­nents of the sup­ply chain,” Hulme says. 

Com­pa­nies can use the data gen­er­at­ed by the dig­i­tal twin to antic­i­pate prob­lems well before they strike. They can even apply it to inform their mar­ket­ing cam­paigns, accord­ing to Hyde. 

“Retail­ers need to be proac­tive about using pro­mo­tions to con­trol the sales process, both to max­imise sales and to min­imise the risk of ful­fil­ment delays caused by exces­sive demand,” he says. 

For instance, pro­mo­tions on slow-mov­ing prod­uct lines should be pri­ori­tised over pop­u­lar lines to pre­vent over­stock­ing and under­stock­ing. At the same time, “pro­mot­ing prod­ucts that aren’t yet in the ware­house, or even in the coun­try, is a fast track to cus­tomer dis­ap­point­ment,” Hyde warns.

While busi­ness­es are gen­er­al­ly aware of AI’s poten­tial ben­e­fits, some are strug­gling to use it opti­mal­ly. That’s the key find­ing of a sur­vey of 350 ware­house man­agers in the UK and the US by Lucas Sys­tems, a spe­cial­ist in ware­hous­ing tech­nol­o­gy. The con­sen­sus among the respon­dents was that AI could give them a return on invest­ment of 60% with­in five years of its adop­tion. Yet 99% admit­ted that they were find­ing it hard to make the best use of the tech­nol­o­gy, while 90% said that they need­ed more sup­port with imple­ment­ing it. 

“There is a huge mis­un­der­stand­ing about the capa­bil­i­ty of AI, which leads to so many com­pa­nies get­ting this wrong,” Hulme says.

One com­mon mis­con­cep­tion that com­pa­nies have about AI is that it’s cost­ly and involves a high­ly tech­ni­cal set-up, he adds. Anoth­er is that it’s only tru­ly ben­e­fi­cial to com­pa­nies han­dling a high vol­ume of prod­ucts. 

Hyde believes there’s nev­er been a more appro­pri­ate time for retail­ers and ful­fil­ment com­pa­nies to start using AI. Prepar­ing for the fes­tive peri­od is always a chal­lenge but, giv­en the well-report­ed lack of HGV dri­vers on top of a gen­er­al staff short­age this year, busi­ness­es will find it hard­er than usu­al to cov­er any spike in ware­house demand. 

“Vague promis­es aren’t good enough,” Hyde stress­es. Cus­tomers expect imme­di­ate ser­vice, but this won’t be achiev­able with the cur­rent bot­tle­necks. By using AI-pow­ered tools to keep on top of inven­to­ry, retail­ers can make real­is­tic deliv­ery pledges, he notes. Cus­tomers would rather receive a deliv­ery with­in an agreed time slot than be offered a next-day deliv­ery ser­vice that’s not fea­si­ble. 

Retail­ers should be ask­ing them­selves how quick­ly ware­hous­es can respond to cus­tomer demand and get prod­ucts dis­patched, accord­ing to Hyde. And key to this is know­ing exact­ly what’s in stock at any giv­en point and which prod­uct lines need to be pro­mot­ed.

“Care­ful plan­ning is vital if retail­ers are to get the most from peak trad­ing sea­son and, cru­cial­ly, keep their promis­es,” he con­cludes. “Retail­ers need to know at all times what they have and when they can deliv­er it.”