Why AI? Why now?

Imam Hoque, Chief oper­at­ing offi­cer and head of prod­uct, Quan­texa

Just as the Indus­tri­al Rev­o­lu­tion trans­formed the world dur­ing the 18^th and 19^th cen­turies, we are fac­ing the dawn of an equal­ly far-reach­ing arti­fi­cial intel­li­gence or AI rev­o­lu­tion that will be mea­sured in years rather than decades. AI has reached the point where it is capa­ble of sur­pass­ing the deci­sion-mak­ing of humans in many sit­u­a­tions; con­sis­tent­ly, accu­rate­ly, 24/7 and based on more facts. But why now and how do busi­ness­es har­ness this capa­bil­i­ty?

AI has been around since the 1960s; only now is the con­flu­ence of three key fac­tors com­ing togeth­er. First­ly, AI has lan­guished as a “dis­em­bod­ied brain in a jar”, iso­lat­ed from the real world. Today we have reached the point where dig­i­tal chan­nels are becom­ing the norm, pro­vid­ing the brain with all the sens­es and limbs it needs. Our cus­tomers and sup­pli­ers com­mu­ni­cate with us elec­tron­i­cal­ly, so why place a human in the loop to make deci­sions?

Sec­ond is the avail­abil­i­ty of data, the new “dig­i­tal fuel”. Data lakes are pop­ping up every­where, appli­ca­tion pro­gram­ming inter­faces or APIs are avail­able on the inter­net to access all man­ner of data, and we as indi­vid­u­als and busi­ness­es are gen­er­at­ing a huge but insight­ful “dig­i­tal exhaust”.

Final­ly, where­as steam machines were fash­ioned out of steel with large cap­i­tal invest­ment, huge cloud com­pute capac­i­ty can be pro­vi­sioned in sec­onds for dol­lars. The AI rev­o­lu­tion is set to take the ser­vice indus­try by storm, with the poten­tial to change the role of the white-col­lar work­er irre­versibly.

OPPORTUNITY FOR THOSE WHO RACE TO EMBRACE AI

This is a tru­ly glob­al race; the oppor­tu­ni­ty is automat­ing or aug­ment­ing many deci­sion-mak­ing tasks in real time, more accu­rate­ly using more facts. The start­ing gun has been fired, there is a pack out front – the glob­al inter­net play­ers: Ama­zon, Pay­Pal, Google, Face­book and friends. The stakes are fur­ther com­pound­ed by gov­ern­ment reg­u­la­tion dri­ving com­pe­ti­tion through open bank­ing and the revised Pay­ment Ser­vices Direc­tive.

If the glob­al inter­net brands steal a march, with their low cost-bases, seam­less real-time inter­ac­tion and high­ly com­pet­i­tive­ly priced prod­ucts, it could prove hard to catch them.

AI can gen­er­ate sig­nif­i­cant insight into your cus­tomers, sup­pli­ers and busi­ness rela­tion­ships from huge vol­umes of seem­ing­ly dis­parate sources of data. This under­pins the abil­i­ty to make auto­mat­ed unique deci­sions per indi­vid­ual or organ­i­sa­tion on a whole range of top­ics. It can also iden­ti­fy hid­den trends and pat­terns of behav­iour.

The appli­ca­tions are poten­tial­ly lim­it­less and have a pro­found impact on head­count as well as oppor­tu­ni­ties for more inno­v­a­tive offer­ings or sig­nif­i­cant­ly improved cus­tomer ser­vice.

CASE STUDIES

Imag­ine being able to cre­ate 80 per cent more, tar­get­ed, poten­tial cus­tomer prospects and com­plete the pre-qual­i­fi­ca­tion auto­mat­i­cal­ly. By com­bin­ing glob­al lists of busi­ness­es, direc­tors, investors and share­hold­ers, along with your own cur­rent cus­tomer base, AI sys­tems can do just that. They will work out what your best cus­tomers look like, scour the glob­al lists, find sim­i­lar busi­ness­es, rate them based on a whole range of fac­tors, present you the best oppor­tu­ni­ties and even go as far as telling you which of your cur­rent cus­tomers could make an intro­duc­tion.

The AI rev­o­lu­tion is set to take the ser­vice indus­try by storm, with the poten­tial to change the role of the white-col­lar work­er irre­versibly

The cost of com­pli­ance is a huge bur­den, with finan­cial insti­tu­tions deploy­ing tens of thou­sands of staff to per­form know-your-cus­tomer, sanc­tions checks and anti-mon­ey laun­der­ing trans­ac­tion mon­i­tor­ing, react­ing to adverse media or Pana­ma leaks. By pro­duc­ing a full con­tex­tu­al view of prospects and cus­tomers, link­ing data across many sources, AI sys­tems can auto-clas­si­fy for poten­tial crim­i­nal­i­ty far more accu­rate­ly than tra­di­tion­al rule-based sys­tems or humans fol­low­ing rules.

Iden­ti­fy­ing more illic­it mon­ey move­ments, and by improv­ing the 98 per cent false pos­i­tive rates, AI can save up to 70 per cent of staff effort. This is crit­i­cal to mak­ing a real impact on the pro­ceeds of crime, reduc­ing organ­ised crime, human traf­fick­ing, ter­ror, rad­i­cal­i­sa­tion, inequal­i­ty through tax eva­sion and cor­rup­tion.

Pre­dict­ing risk of default more accu­rate­ly is pos­si­ble when AI is pro­vid­ed more con­text about a busi­ness being assessed. After all, you would not buy a house by look­ing through the let­ter box; you would go inside, look around the neigh­bour­hood and then make your deci­sion. Like­wise, using share­hold­er and direc­tor rela­tion­ships, even the busi­ness­es’ own cus­tomers and sup­pli­ers net­worked togeth­er will dri­ve supe­ri­or AI deci­sion out­comes.

SECRET TO AI SUCCESS

Step 1: Get­ting your data pre­pared
Organ­i­sa­tions under­stand that data cre­ates busi­ness and cus­tomer insight, and their response has been to set up a chief data office and data lakes. Not all have seen the val­ue, how­ev­er, as they have not yet plumbed in auto­mat­ed AI deci­sion-mak­ing. Peo­ple also pan­ic about data qual­i­ty. Don’t, it will nev­er be per­fect; AI is a game of sta­tis­tics and you can use quan­ti­ty of data to over­come qual­i­ty. Start by mak­ing sure you gen­er­ate a sta­tis­ti­cal sin­gle view of a cus­tomer, don’t wait for some mas­ter data man­age­ment pro­gramme to com­plete as you’ll miss the start­ing gun. Con­text is crit­i­cal; the AI engine needs to be fed with net­worked data pro­vid­ing the full pic­ture to make deci­sions. For exam­ple, why assess a claim in iso­la­tion, when look­ing at the net­work of con­nect­ed claims would prove it is gen­uine or an organ­ised fraud ring. Final­ly, ensure you can gen­er­ate sin­gle cus­tomer views and net­works in real time, not just old-fash­ioned batch.

Step 2: Under­stand­ing AI
Don’t con­fuse AI with robot­ic process automa­tion, which is a basic capa­bil­i­ty that intro­duces anoth­er com­put­er to try to auto­mate exist­ing lega­cy com­put­ers by fol­low­ing the same rules a human oper­a­tor would. AI is a step-change to achieve supe­ri­or deci­sion­ing based on deep­er insight and more data. Don’t only think AI is about machine or deep-learn­ing; AI is more effec­tive as a com­bi­na­tion of tech­niques, effec­tive­ly cre­at­ing an expert sys­tem.

Step 3: Find the right prob­lems
Organ­i­sa­tions that have strug­gled with AI have suc­cumbed to one of an unsuit­able prob­lem, the data was not pre­pared cor­rect­ly or users were alien­at­ed – “com­put­er says no” syn­drome. It is very impor­tant to select care­ful­ly the prob­lems that make a real dif­fer­ence and ensure the right data is avail­able. To solve the user-inter­ac­tion prob­lem, don’t just think of AI as pure­ly machine-learn­ing – a yes/no answer. Engage users to derive expert sys­tem-type rules and mod­els. Then ensure the inter­face presents the full pic­ture and rea­sons under­pin­ning the deci­sion back to the user, effec­tive­ly ampli­fy­ing your best deci­sion-mak­ers.

Step 4: A well-struc­tured pro­gramme
You need streams of work to under­pin a series of AI pilots and projects – data acqui­si­tion, data sci­ences skills, open tech­nol­o­gy data-lake envi­ron­ments, IT engage­ment, pilot-to-pro­duc­tion process, aware­ness and com­mu­ni­cat­ing suc­cess. Don’t wait for the race to be won, make a start today, take baby steps and reap the ben­e­fits of suc­cess.

For more on sin­gle cus­tomer view, con­text, net­works and AI-dri­ven deci­sions please email [email protected] or vis­it www.quantexa.com

OPPORTUNITY FOR THOSE WHO RACE TO EMBRACE AI

CASE STUDIES