Not just for big business: how AI went mainstream

Not so long ago, AI was the pre­serve of the largest organ­i­sa­tions, main­ly because of its cost and com­plex­i­ty. But this is start­ing to change. As the tech­nol­o­gy becomes more afford­able, the largest host­ing providers, such as Microsoft, Ama­zon and Google, are open­ing up access to shared resources and pre-pack­aged AI sys­tems with offer­ings aimed at small­er busi­ness­es.

With AI becom­ing sophis­ti­cat­ed enough to pro­gram itself, some lead­ing tech­nol­o­gy providers are even delv­ing into the world of ‘cit­i­zen devel­op­ers’, as David Shri­er, pro­fes­sor of prac­tice, AI and inno­va­tion at Impe­r­i­al Col­lege Busi­ness School, explains.

“This capa­bil­i­ty is grow­ing clos­er. Under such a mod­el, a small busi­ness own­er would rent AI capac­i­ty from a large tech com­pa­ny and describe a prob­lem ver­bal­ly to the AI. The com­put­er would then write a pro­gram for itself to solve that prob­lem,” he says.

In the mean­time, SMEs can start to ben­e­fit from the AI rev­o­lu­tion by increas­ing their knowl­edge of the tech­nol­o­gy and under­stand­ing its under­pin­ning prin­ci­ples. So-called low-code or even no-code AI sys­tems are also avail­able for use by those with no pro­gram­ming knowl­edge.

“Small busi­ness own­ers can become smarter about these tech­nolo­gies through online cours­es that pro­vide insights into the field. They should also con­sid­er hir­ing an expert con­sul­tant,” Shri­er says.

Small­er firms can adopt var­i­ous tech­nolo­gies under the AI umbrel­la to enhance a range of oper­a­tions. They could use chat­bots pow­ered by an expert sys­tem to cut their cus­tomer-ser­vice costs, for instance, or they could apply machine-learn­ing algo­rithms to opti­mise their dig­i­tal mar­ket­ing expen­di­ture.

Key fac­tors to con­sid­er when eval­u­at­ing AI sys­tems include cost, ease of use and scal­a­bil­i­ty. A small­er busi­ness will gen­er­al­ly pre­fer a sys­tem that can grow with the com­pa­ny as its needs evolve and one that requires users to under­take min­i­mal train­ing. The most cost-effec­tive way to start with AI is by apply­ing it to rel­a­tive­ly sim­ple tasks on which it can have a big impact. This way, the firm is like­ly to achieve a faster return on its invest­ment.

Found­ed in 2005, Neom Organ­ics is a cos­met­ics com­pa­ny that employs about 150 peo­ple. The Har­ro­gate-based firm is increas­ing its use of AI to auto­mate numer­ous man­u­al process­es and improve their effi­cien­cy. But its over­rid­ing goal is to improve the cus­tomer expe­ri­ence it offers.

The com­pa­ny has devel­oped a sophis­ti­cat­ed tech­no­log­i­cal infra­struc­ture that it calls Neom IQ. This gath­ers data from across the busi­ness and uses it to pow­er var­i­ous AI-dri­ven mod­els, includ­ing stock demand fore­casts, finan­cial per­for­mance pro­jec­tions and the cre­ation of per­son­alised prod­uct rec­om­men­da­tions for online cus­tomers.

The com­pa­ny has worked with a cou­ple of AI con­sul­tants to “build out the ecosys­tem” and is now bring­ing in more experts in data mod­el­ling and machine learn­ing, accord­ing to its CEO, Oliv­er Men­nell.

“Once the sys­tem is set up, it isn’t com­plex to run, but it does rely on hav­ing best-in-class data integri­ty,” he says. “We work with some of the best tech­nolo­gies for our data ware­hous­ing, report­ing and mar­ket­ing capa­bil­i­ties, but have devel­oped our own pre­dic­tive mod­el­ling and per­son­al­i­sa­tion tools in house.”

SME lead­ers don’t need high-lev­el tech skills to adopt AI sys­tems, but they do at least need to under­stand the basic ter­mi­nol­o­gy, the kinds of prob­lems the tech­nol­o­gy is good at solv­ing and the type of con­sul­tant they’ll want to hire to help them imple­ment it.
Men­nell adds that it’s “also crit­i­cal to have a tech­nol­o­gy strat­e­gy and over­sight of how these sys­tems are being used in your busi­ness, because they will ulti­mate­ly dis­rupt most, if not all, indus­tries”.

Although sev­er­al sophis­ti­cat­ed and cost­ly AI sys­tems were avail­able to Neom, it chose a rel­a­tive­ly cheap, sim­ple and adapt­able option. This enabled mem­bers of Mennell’s team to become more hands-on with the data and build their own mod­els.

“As we devel­op, we are adding more and more com­plex data sets and sources into the sys­tem,” he says. “We’re find­ing more inno­v­a­tive ways to use this data to improve the cus­tomer expe­ri­ence.”

By analysing huge amounts of data, detect­ing trends as they devel­op and rec­om­mend­ing poten­tial respons­es, AI offers small­er firms a much-improved abil­i­ty to fore­cast new areas of demand – a key fac­tor in fuelling their growth. But any SME leader who’s plan­ning to bring AI into their com­pa­ny first needs to be aware of a cou­ple of impor­tant caveats.

A key chal­lenge for them is to ensure that any inher­ent bias, based on fac­tors such as age, race and gen­der, is elim­i­nat­ed. They must also sat­is­fy them­selves that the peo­ple cre­at­ing the tech­nol­o­gy have not embed­ded their own prej­u­dices with­in it.

Will Rich­mond-Cog­gan is a direc­tor at Freeths, a top-50 UK law firm by rev­enue that spe­cialis­es in tech­nol­o­gy and pri­va­cy issues. He advis­es SME lead­ers to “find out how the AI has been trained and what steps have been tak­en to elim­i­nate or con­trol for inher­ent bias in any train­ing data set. You will need exper­tise to under­stand the answer. Such con­cerns can best be allayed by prop­er due dili­gence when you’re mak­ing pur­chas­ing deci­sions.”

Data pro­tec­tion can be anoth­er area of con­cern for SMEs. The spe­cif­ic rules with­in the Gen­er­al Data Pro­tec­tion Reg­u­la­tion that gov­ern auto­mat­ed deci­sion-mak­ing require addi­tion­al safe­guards. This means that small­er busi­ness­es will need to give more thought to how they inform employ­ees and cus­tomers about what they are doing if they adopt AI sys­tems.