SAP recent­ly extend­ed the capa­bil­i­ties of SAP Data­s­phere and SAP Ana­lyt­ics Cloud, enhanc­ing two of its major data ana­lyt­ics solu­tions with gen­er­a­tive AI, data gov­er­nance, and knowl­edge graph­ing capa­bil­i­ties in its efforts to estab­lish a uni­fied busi­ness data fab­ric for customers. 

As detailed dur­ing the SAP Data Unleashed vir­tu­al event, new data mod­el­ing capa­bil­i­ties with­in SAP Data­s­phere, along with vec­tor capa­bil­i­ties in SAP HANA Cloud, are gen­er­al­ly avail­able to improve the data service’s inter­ac­tions with large lan­guage mod­els (LLMs), sup­port­ing gen­er­a­tive-AI out­puts with busi­ness con­text and inhibit­ing AI-induced data hallucinations.

To dive deep­er into the announce­ments and reflect on SAP’s larg­er data strat­e­gy, ASUG con­nect­ed with Irfan Khan, Pres­i­dent & Chief Prod­uct Offi­cer, SAP HANA Data­base & Analytics. 

A 12-year vet­er­an of SAP, Khan cur­rent­ly leads glob­al devel­op­ment, prod­uct, and solu­tion man­age­ment for the company’s entire data­base, data man­age­ment, and ana­lyt­ics port­fo­lio, which includes SAP HANA, SAP Ana­lyt­ics Cloud, and SAP Data­s­phere. Pri­or to assum­ing this role in 2021, Khan spent six years in sales, most promi­nent­ly as the pres­i­dent and chief rev­enue offi­cer for the SAP Plat­form & Tech­nolo­gies orga­ni­za­tion, where he man­aged glob­al sales and go-to-mar­ket (GTM) for all SAP data­base assets.

Below, Khan dis­cuss­es the crit­i­cal role that con­nect­ing enter­prise data will play in enabling SAP cus­tomers to lever­age gen­er­a­tive AI, the company’s vision for extend­ed plan­ning and analy­sis (xP&A), and the uni­fy­ing role of the busi­ness data fab­ric in SAP’s inno­va­tion roadmap.

This inter­view has been edit­ed and condensed.

ASUG: Let’s set the scene. A year ago, SAP announced SAP Data­s­phere and detailed its impor­tance with­in the idea of archi­tect­ing a ​“busi­ness data fab­ric” for cus­tomers. Can you recap that recent direc­tion for our readers?

Irfan Khan: SAP Data­s­phere is the SAP man­aged ser­vice, part of SAP Busi­ness Tech­nol­o­gy Plat­form. It’s there as a means of help­ing cus­tomers nav­i­gate the high­ly diverse and equal­ly so het­ero­ge­neous envi­ron­ments and datasets they have to inter­act with. We’ve phrased this ter­mi­nol­o­gy of the ​“busi­ness data fab­ric” as a means of archi­tec­tural­ly shar­ing clar­i­ty around how we achieve our goals for our customers. 

Our first goal: no cus­tomer should be left behind, and no data should be left behind. It doesn’t mat­ter if you start off with an on-prem envi­ron­ment, or for that mat­ter in a pri­vate cloud or pub­lic cloud envi­ron­ment. All of your data should be acces­si­ble to you. As we look at more gen­er­a­tive-AI use cas­es, the last thing we’d want to do is to mar­gin­al­ize the val­ue of some data because it’s inac­ces­si­ble or because our cus­tomers haven’t fig­ured out how to pipeline the data into their main processes. 

How does SAP Ana­lyt­ics Cloud fit in? Among the announce­ments, SAP promised improved inte­gra­tion between SAP Data­s­phere and SAP Ana­lyt­ics Cloud, which will also pro­vide users with access to gen­er­a­tive-AI assis­tant Joule. Walk us through the inte­gra­tions at play here.

Khan: In pro­vid­ing a rich­er expe­ri­ence for Joule, we will embed Joule direct­ly with­in SAP Ana­lyt­ics Cloud, so that will serve as our foun­da­tion­al start­ing point. SAP is one of the largest plan­ning ven­dors in the mar­ket — not that you would know that, because we have a very dif­fused plan­ning foun­da­tion. You have sup­ply chain plan­ning with SAP Inte­grat­ed Busi­ness Plan­ning (IBP), work­force plan­ning in SAP Suc­cess­Fac­tors, and ter­ri­to­ry plan­ning in SAP Sales Cloud, for­mer­ly Cal­lidus­Cloud. SAP has many dif­fer­ent plan­ning capabilities. 

Whilst there’s val­ue in plan­ning, and whilst there’s val­ue for users who’ve been using those explic­it plan­ning capa­bil­i­ties, there’s more ben­e­fit if you have seam­less­ly inte­grat­ed end-to-end plan­ning. This is the direc­tion that we’ve been fol­low­ing; mar­ket com­men­ta­tors cat­e­go­rize it under the ban­ner of extend­ed plan­ning and ana­lyt­ics (xP&A).

As a plan­ning foun­da­tion, SAC in con­junc­tion with SAP Data­s­phere pro­vides our abil­i­ty to ser­vice the needs of xP&A. It does so in two ways. First, SAP Data­s­phere gives you access to all the dif­fer­ent plan­ning con­tent. Sup­ply chain plan­ning in IBP, for exam­ple, runs on SAP HANA Cloud, which is a foun­da­tion of SAP Data­s­phere. Now we have the fed­er­a­tion capa­bil­i­ty from SAP HANA to SAP HANA, through SAP HANA Cloud run­ning under­neath IBP for sup­ply chain plan­ning. As a con­se­quence, SAP Data­s­phere is able to con­sume plan­ning mod­els direct­ly. We can do a seman­ti­cal onboard­ing of the dif­fer­ent plan­ning con­tent that comes from those dif­fer­ent plan­ning solu­tions. That is, in itself, a huge value-add. 

Add to that the new SAP Data­s­phere knowl­edge graph, and add Joule, and think about the cor­re­la­tion of those tech­nolo­gies with SAP Ana­lyt­ics Cloud as a plan­ning foun­da­tion, which now has the abil­i­ty to estab­lish extend­ed plan­ning across all the dif­fer­ent plan­ning con­tent across SAP. Now, with the knowl­edge graph, you can ask open-end­ed ques­tions of that foundation. 

If you’re a plan­ner, and you want to go through a process and look for a fore­cast or pre­dic­tive mod­el, and you want to asso­ciate that with a plan­ning foun­da­tion, that’s with­in your abil­i­ty. If you’re a mar­ke­teer with 2 mil­lion euros to spend and you want to make the most sig­nif­i­cant impact with that 2 mil­lion, you can use a plan­ning foun­da­tion to deter­mine which is your most sig­nif­i­cant ter­ri­to­ry in terms of rev­enue. You could stip­u­late the ter­ri­to­ry has to have a demo­graph­ic of over a mil­lion poten­tial con­sumers, and you want to be able to con­vert that. Say you’re run­ning a trade pro­mo­tion; you’ve got to plan for that. Around sup­ply chain plan­ning, you’ve got to be able to fig­ure out the finance impli­ca­tions. Per­haps you need to have trade pro­mo­tions, where you incen­tivize cer­tain sell­ers. This is about how you tie all those dif­fer­ent areas together. 

That’s a com­mon sce­nario many SAP cus­tomers run today. They’re chal­lenged, because each and every one of those expe­ri­ences will result in a large amount of data move­ment, build­ing out plan­ning mod­els, inte­grat­ing plan­ning mod­els, and build­ing a user expe­ri­ence on top of them. Joule is avail­able through SAP Ana­lyt­ics Cloud, which inte­grates with SAP Data­s­phere, which in turn now offers knowl­edge graph­ing, which allows you to access a vari­ety of dif­fer­ent datasets and har­mo­nize that data in one seman­ti­cal­ly enriched data lay­er, served up through SAP Data­s­phere. This allows you to accel­er­ate those types of use cas­es, also using a com­bi­na­tion of SAP BTP ser­vices on top of that.

ASUG: To be clear, xP&A capa­bil­i­ty is not replac­ing SAP Inte­grat­ed Busi­ness Plan­ning (IBP) but rather, as the name sug­gests, extend­ing its reach. 

Khan: That’s right. We wouldn’t for a moment con­sid­er elim­i­nat­ing SAP IBP in this process, because that’s not in the best inter­ests of the cus­tomer. You use IBP for syn­chro­nized plan­ning, and you look at it from the man­u­fac­tur­ing process side. There is an excel­lence in the sup­ply chain that IBP encap­su­lates. But what you do want to do is use the mod­els that you’re using in sce­nar­ios around sup­ply chain plan­ning to extend their [con­nec­tiv­i­ty] to your finan­cial or work­force plan­ning. That’s how we look at this whole evo­lu­tion: extend­ing P&A, so that you can run ana­lyt­ics on top of plan­ning con­tent, to do forecasting.

Think of a con­troller that comes up with a bud­get. If you want to have the plan­ner decide where the plan­ning con­tent needs to go, plan­ning val­ues by mar­ket unit or by ter­ri­to­ry, and then you want to run com­plex pre­dic­tive ana­lyt­ics on top of that, this links all those pieces togeth­er. What was miss­ing before was that sin­gle, seman­ti­cal­ly rich har­mo­niza­tion lay­er, aka SAP Data­s­phere, with its notion of a busi­ness data fab­ric archi­tec­ture that allows us to play to the strength of the entire cus­tomer land­scape, instead of bias­ing it by what you can con­trol because you can only access sub­sec­tions of data. 

With gen­er­a­tive AI sweep­ing across enter­prise tech­nol­o­gy, data qual­i­ty and secu­ri­ty is top of mind for our mem­bers, as is trans­paren­cy as to how AI inter­acts with enter­prise data to ensure they can trust and val­i­date the plan­ning and analy­sis it generates.

Khan: One of the announce­ments we made involved extend­ing our rela­tion­ship and part­ner­ship with Col­li­bra around gen­er­a­tive-AI gov­er­nance. That plays to what you said a moment ago: if you want to ensure you’re look­ing at data of the right qual­i­ty, with the right seman­tic enrich­ment that you need, you need to have a lev­el of gov­er­nance around that. We don’t want hal­lu­ci­na­tions in these large lan­guage mod­els. We’re approach­ing that with real prag­ma­tism, con­sid­er­ing how we can use the best of breed to be able to extend the val­ue of SAP. Where we see the best of breed accel­er­at­ing ben­e­fits, we want to cap­i­tal­ize on that. 

Sim­i­lar­ly, but inter­nal­ly, we also have our AI ethics approach. With SAP’s intrin­sic under­stand­ing of cus­tomer data and busi­ness process­es, we can under­stand the con­text of data, rather than tak­ing it out of con­text. In HR sys­tems, if you’re look­ing at employ­ee demo­graph­ics, and you aren’t look­ing at it through a his­tor­i­cal lens, per­haps there’s been a huge push, based upon cer­tain KPIs that larg­er orga­ni­za­tions have, that adds con­text to the demo­graph­ics. Envi­ron­men­tal and social gov­er­nance (ESG) falls into that as well. You look at the green bal­ance sheet, where cus­tomers now want to be able to move the posi­tion­ing of how eco­log­i­cal­ly they’re pro­gress­ing in terms of car­bon neu­tral­i­ty. You need to look at the broad­er con­text of that data. This is why, when we’re look­ing at AI, we’re look­ing at busi­ness context.

Joule is able to look at data with that same eth­i­cal lens. And you have to look through that, because oth­er­wise you’re going to be ask­ing ques­tions that are either out­side the realm of what would be con­sid­ered to be eth­i­cal or out of con­text of the kinds of ques­tions that need to be asked. We view this as one elab­o­rate but ele­gant way of putting all these pieces togeth­er: elab­o­rate because there are lots of mov­ing parts, ele­gant because of the way we can pre­cise­ly con­nect them using the busi­ness data fabric. 

The val­ue of us inte­grat­ing Joule across all of our dif­fer­ent lines of busi­ness, along with our gen­er­a­tive AI hub that gives you access to a broad range of LLMs, and using the foun­da­tions of SAP BTP — with its SAP Build Code capa­bil­i­ties also unlock­ing pro-code devel­op­ment capa­bil­i­ties—is that you can use gen­er­a­tive-AI with Joule to cre­ate the next gen­er­a­tive-AI appli­ca­tions, build new data mod­els, and even build test infra­struc­ture around all of that. It real­ly builds an end-to-end sto­ry for our cus­tomers, from the pro-code devel­op­ment side to the con­sump­tion side for plan­ning capa­bil­i­ties, all the way to hav­ing a sin­gle dig­i­tal-copi­lot expe­ri­ence across all of your SAP applications. 

SAP wants to empow­er both busi­ness and tech­ni­cal users to suc­cess­ful­ly nav­i­gate its solu­tions. What capa­bil­i­ties will these data-cen­tric announce­ments help to build, for cus­tomers look­ing to ensure those in IT and on the busi­ness side of oper­a­tions have equal con­fi­dence to make inquiries with­in SAP systems?

Khan: There is def­i­nite­ly a spec­trum of dif­fer­ent busi­ness users and tech­ni­cal users. And we can’t obfus­cate one against the oth­er. Take SAP Data­s­phere — that’s appeal­ing to data archi­tects and data pro­fes­sion­als who need to build inte­gra­tion points for SAP and non-SAP data. One of the huge move­ments that we’ve been pur­su­ing inter­nal­ly, across all of our dif­fer­ent lines of busi­ness, con­cerns the abil­i­ty for each line of busi­ness to cre­ate data products. 

What is a data prod­uct? Essen­tial­ly, it’s an encap­su­la­tion of a seman­ti­cal­ly enriched pay­load of data. A good exam­ple would be an invoice, right? An invoice is a seman­ti­cal­ly enriched dataset that has cus­tomer data, order infor­ma­tion, line-item infor­ma­tion, and VAT infor­ma­tion; in aggre­gate, that con­sti­tutes an invoice. It’s not the raw data or any of those indi­vid­ual con­stituent parts that cus­tomers care about. They care about an invoice, and its sup­ply chain; you care about an order and inven­to­ry. All of those things need to be reflect­ed from the appli­ca­tion side as data prod­ucts. SAP Data­s­phere can con­sume those data prod­ucts. It’s a con­sumer of data prod­ucts by lines of busi­ness of SAP. 

Imag­ine that you’ve got SAP Suc­cess­Fac­tors. And you cre­ate, for instance, spe­cif­ic data prod­ucts around the onboard­ing expe­ri­ence. Or, in SAP Ari­ba, it could be based around the pro­cure­ment expe­ri­ence with­in a net­work, around sup­pli­er spend analy­sis, for exam­ple. All of this infor­ma­tion is dis­parate­ly asso­ci­at­ed with each of the dif­fer­ent lines of busi­ness. And you, as an SAP cus­tomer, could have mul­ti­ple SAP prop­er­ties. Maybe you have core ERP, in SAP S/4HANA, you’ve got SAP Ari­ba, and you’ve got SAP SuccessFactors. 

How is it that you can com­bine all that data togeth­er in a seam­less expe­ri­ence? As a busi­ness user, you’re rely­ing upon IT to some­how stitch all that togeth­er for you: typ­i­cal­ly, to inte­grate it all togeth­er, put it in a data lake, and build it out in terms of a new data mod­el. But the val­ue of SAP Data­s­phere, of hav­ing these data prod­ucts served up via dif­fer­ent lines of busi­ness, is that we can seman­ti­cal­ly onboard those data prod­ucts into SAP Data­s­phere. And by doing so, your data pro­fes­sion­al only needs to under­stand an inven­to­ry, which includes all the dif­fer­ent data prod­ucts out there. 

Auto­mat­i­cal­ly, you’re build­ing a real foun­da­tion, hav­ing that busi­ness data fab­ric. This fab­ric is able to pro­vide you with con­nec­tiv­i­ty to data that ordi­nar­i­ly would involve a tech­ni­cal lift-and-shift or inte­gra­tion project. As a busi­ness user, you can inter­ro­gate those spe­cif­ic data prod­ucts. You can take a look at build­ing those plan­ning use cas­es but not hav­ing to go through IT to curate, pro­vi­sion, and pull all this data from all those dif­fer­ent envi­ron­ments. That is the ben­e­fit and the beau­ty of the entire end-to-end SAP data strat­e­gy com­ing to life. 

Data prod­ucts curat­ed by dif­fer­ent lines of busi­ness are ren­dered and made avail­able through a cat­a­log, which is what SAP Data­s­phere is, reg­is­ter­ing those dif­fer­ent data prod­ucts. You can seman­ti­cal­ly onboard your Busi­nes­sOb­jects envi­ron­ment, objects with­in data in SAP Busi­ness Ware­house, to seman­ti­cal­ly onboard them or go to your SAP ERP Cen­tral Com­po­nent (ECC) where your Core Data Ser­vices (CDS) Views live. You can bring all that data togeth­er in one sin­gle loca­tion, in a vir­tu­al­ized way or in a phys­i­cal way. As a busi­ness user or as a tech­nol­o­gy user, we’re blur­ring the lines, and tak­ing away a lot of the heavy lift­ing that ordi­nar­i­ly would have been necessary.

At last fall’s SAP TechEd announce­ments, vec­tor capa­bil­i­ties for SAP HANA Cloud were announced to enhance the enterprise’s inter­ac­tions with LLMs and by exten­sion its abil­i­ty to lever­age gen­er­a­tive AI. Since then, how has the con­ver­sa­tion about vec­tor capa­bil­i­ties in SAP HANA Cloud pro­gressed at SAP?

Khan: There’s a bunch of mov­ing parts. You’ve got the vec­tor­ing capa­bil­i­ty in SAP HANA Cloud, knowl­edge graphs sit­ting inside of SAP Data­s­phere, Joule’s inte­gra­tion with SAP Ana­lyt­ics Cloud to pro­vide a more copi­lot-based expe­ri­ence… All of these areas are correlated. 

I’d say, with­out hes­i­ta­tion, that SAP HANA has been a major vehi­cle for inno­va­tion for a num­ber of years. We talked about this in terms of the SAP HANA Cloud evo­lu­tion, split­ting out the stor­age and com­pute lay­er; that’s now part of the foun­da­tion of SAP HANA as it has mod­ern­ized for the cloud-native world. When announce­ments were first made around Ope­nAI and Chat­G­PT, it became very quick­ly evi­dent that LLMs, from a text-based per­spec­tive, are exceed­ing­ly good at ask­ing two-dimen­sion­al types of ques­tions where you’ve encap­su­lat­ed the data and infor­ma­tion need­ed to gen­er­ate a text response. When you get into those more spe­cif­ic, con­tex­tu­al busi­ness ques­tions — for exam­ple, if you were to ask a ques­tion on val­ue-added tax (VAT) har­mo­niza­tion in mul­ti­ple juris­dic­tions — a lot of busi­ness con­text needs to be known.

Hav­ing a vec­tor capa­bil­i­ty, you can go through that drag-and-drop frame­work process, iter­ate through, and be able to pro­vide that busi­ness con­text to an LLM, know­ing and trust­ing that it’s com­ing from SAP. That has orders of mag­ni­tude more impor­tance and val­ue to the end cus­tomer, and it moves away from mass hal­lu­ci­na­tions that you get with LLMs that are trained pure­ly on his­tor­i­cal, pub­lic data from the Inter­net. Because LLMs can build you a mod­el that will allow you to fig­ure out what the VAT thresh­olds should be, but it does­n’t give you the answer. It gives you a means to solve the prob­lem. Cus­tomers don’t want to do that. They want an answer, and they want it to be inject­ed and infused with that exact lev­el of val­ue in context. 

If you can sum­ma­rize data and dri­ve actions based upon busi­ness con­text, that’s a val­ue-add. We’re see­ing that our cus­tomers want to look at gen­er­a­tive AI not just in an arti­fi­cial aca­d­e­m­ic sense, but real­ly plug it into busi­ness con­text. That’s why the whole busi­ness AI move­ment of SAP goes along with these data announce­ments. They’re insep­a­ra­ble. You won’t find busi­ness AI or any gen­er­a­tive AI being suc­cess­ful with­out a very sub­stan­tial data position. 

For more cov­er­age of recent SAP devel­op­ments, dive into the SAP Data Unleashed announce­ments and the TechEd 2023 announce­ments that pre­ced­ed them.