Though many orga­ni­za­tions are inter­est­ed in the idea of bring­ing auto­mat­ed intel­li­gence to their work, few have fig­ured out a prac­ti­cal way to do this yet. There’s no doubt that machine learn­ing can aug­ment the pow­er of the human work­force and help us under­stand so much more than we can today with our cur­rent ana­lyt­ics. The hard part now is know­ing what we do not know. But machines, as they get smarter, may be able to tell us. 

The Pow­er of Human + Machine Learning

To get the most out of a pre­dic­tive mod­el, you need to con­stant­ly retrain each mod­el and man­age its end-to-end life cycle. SAP is aim­ing to do this based on rules and pre­de­fined pat­terns that machine learn­ing can iden­ti­fy and apply so the sys­tem can con­stant­ly retrain itself.

For exam­ple, if the finance depart­ment mis­tak­en­ly receives two invoic­es for the same prod­uct, the sys­tem can use machine learn­ing to not only catch this error, but also learn how to find it again. No need to hire a data sci­en­tist or math­e­mati­cian to repro­gram the algo­rithm — the sys­tem learns auto­mat­i­cal­ly for you. This is how we can all train the machine brains to go to work for us, mas­sive­ly expand­ing the speed and scope at which we can cur­rent­ly accom­plish our dai­ly tasks. 

Where humans can bring in their knowl­edge and con­trol is in build­ing the pre­dic­tive mod­els in the first place. You can cre­ate mul­ti­ple mod­els, each based on a spe­cif­ic hypoth­e­sis, and then test them in your sys­tems in real time. As you com­pare them with actu­al data, you can then decide which of the mod­els is the best one to use. 

Back to Machine Brain School

When it comes to machine learn­ing, it’s a con­stant case of back to school if we want these new tools to remain effec­tive. So where will all of this train­ing mate­r­i­al in the form of data come from? 

SAP has a breadth of inter­nal and part­ner use cas­es to draw on to feed the machine brains. SAP explains that it is now work­ing with major enter­prise part­ners, includ­ing Accen­ture, to iden­ti­fy the use cas­es that these machine learn­ing fea­tures will need to draw their intel­li­gence from. This spe­cif­ic work is focused on infor­ma­tion drawn from SAP S/4HANA, SAP S/4HANA Cloud Edi­tion, SAP Hybris, SAP Ari­ba, and more prod­ucts to come. 

How to Learn With­out Per­fect, Clean Data

When build­ing mod­els with­out the help of machine learn­ing tools like SAP’s, you need per­fect, clean data from your enter­prise to run them. If you real­ly want to fore­cast some­thing, you need to start with accu­rate and sta­ble data. Oth­er­wise you’ll find the truth in the old ​“garbage in, garbage out” adage. It’s not real­is­tic for most com­pa­nies to find this type of data in their sys­tems with­out hav­ing to do a lot of prepa­ra­tion to clean it. 

That’s where machine learn­ing pow­ered by real data from real use cas­es comes in. The sys­tem will use rules and pre­de­fined pat­terns to cor­rect the data as it encoun­ters errors, inac­cu­ra­cies, and omis­sions. Mean­while, it takes the best of the real-world data and uses that as a ref­er­ence point as it learns.

Democ­ra­tiz­ing Machine Learn­ing Across Applications

SAP is engi­neer­ing this more-acces­si­ble ver­sion of machine learn­ing intel­li­gence for pre­dic­tive ana­lyt­ics func­tions, avail­able through SAP Pre­dic­tive Ana­lyt­ics App Edi­tion. Its mis­sion is clear: SAP wants to democ­ra­tize machine learn­ing across applications.

SAP Hybris already has machine learn­ing intel­li­gence for pre­dic­tive ana­lyt­ics built in. A pre­vi­ous third-par­ty rec­om­men­da­tion engine has now been replaced with an SAP Hybris engine that is specif­i­cal­ly tuned to com­merce and mar­ket­ing to help with actions such as tar­get­ing and seg­men­ta­tion using the pre­dic­tive mod­el. If users wish to extend its pow­er, then they can do so through SAP Pre­dic­tive Ana­lyt­ics App Edition.

As SAP con­tin­ues to aug­ment SAP Leonar­do with its grow­ing num­ber of indus­try accel­er­a­tors, it will be inter­est­ing to see how task-spe­cif­ic machine learn­ing enhance­ments get smarter in their sub­ject-mat­ter exper­tise. These capa­bil­i­ties could cer­tain­ly find their way into an SAP Leonar­do project, and the SAP Pre­dic­tive Ana­lyt­ics App Edi­tion could fit right with­in a company’s asso­ci­at­ed blueprint.

Build­ing the Machine Brains Right from the Start

If arti­fi­cial intel­li­gence (AI) lives up to its promise of real-world busi­ness impact, then it’s clear that we need to build the machine brains right from the start and make their intel­li­gence as wide­ly acces­si­ble as pos­si­ble through­out an orga­ni­za­tion. If we do this well now, the future will be smarter, in a machine learn­ing kind of way. 

Sign up for our SAP Leonar­do web­cast series to hear more about how you can ben­e­fit from the lat­est advance­ments in machine learn­ing and arti­fi­cial intelligence.