SAP and NVIDIA build­ing safe foun­da­tion for AI agents

AI agents are mov­ing from exper­i­men­ta­tion towards real busi­ness process­es. As they become capa­ble of work­ing with enter­prise appli­ca­tions, data and work­flows, one ques­tion becomes increas­ing­ly impor­tant: how do we make sure they oper­ate secure­ly and with­in the right boundaries?

SAP and NVIDIA are address­ing this chal­lenge through their col­lab­o­ra­tion around NVIDIA Open­Shell, an open-source run­time designed for the secure deploy­ment of autonomous AI agents. SAP engi­neers are con­tribut­ing to Open­Shell, and the tech­nol­o­gy is being embed­ded into the SAP Busi­ness AI Plat­form, mark­ing an impor­tant step towards mak­ing agen­tic AI ready for enter­prise use.

AI that can act also needs boundaries

AI agents can go fur­ther than tra­di­tion­al AI assis­tants. Instead of only pro­vid­ing infor­ma­tion or rec­om­men­da­tions, agents can inter­act with appli­ca­tions and take actions as part of busi­ness process­es. This cre­ates sig­nif­i­cant oppor­tu­ni­ties, but also new require­ments for secu­ri­ty and governance.

NVIDIA Open­Shell pro­vides a run­time bound­ary that con­trols how an agent exe­cutes, what it can access and where infer­ence takes place. At the busi­ness lev­el, Joule Stu­dio run­time applies autho­riza­tion, role-based poli­cies and process con­text before an action reach­es the run­time. Togeth­er, the aim is to enable AI agents to work autonomous­ly while keep­ing their actions con­trolled, gov­erned and auditable.

What does this mean for Islet customers?

For Islet cus­tomers, the real val­ue is not the under­ly­ing tech­nol­o­gy itself, but the busi­ness pos­si­bil­i­ties it can help unlock. AI agents have poten­tial across areas such as finance, pro­cure­ment, sup­ply chain and man­u­fac­tur­ing, where they may need to inter­act with busi­ness-crit­i­cal sys­tems and infor­ma­tion. SAP and NVIDI­A’s work is help­ing estab­lish the secu­ri­ty and gov­er­nance foun­da­tion need­ed for these kinds of use cases.

In prac­tice, this enables more intel­li­gent automa­tion, more effi­cient process­es and new ways of using AI across the busi­ness, while keep­ing per­mis­sions, gov­er­nance and account­abil­i­ty at the cen­tre. For orga­ni­za­tions already run­ning their core process­es with SAP, it also brings agen­tic AI clos­er to where the work and data already are.

From AI exper­i­ments to busi­ness value

The next phase of enter­prise AI is not sim­ply about hav­ing more pow­er­ful AI. It is about con­nect­ing AI secure­ly to the process­es where it can cre­ate real val­ue. The col­lab­o­ra­tion between SAP and NVIDIA is com­bin­ing busi­ness-lev­el gov­er­nance with a secure exe­cu­tion envi­ron­ment for AI agents.

At Islet, we see the oppor­tu­ni­ty in turn­ing tech­no­log­i­cal advances like these into prac­ti­cal busi­ness improve­ments. The impor­tant ques­tion is not where AI can be added, but where it can gen­uine­ly make process­es bet­ter. That is where the jour­ney towards an increas­ing­ly intel­li­gent and autonomous enter­prise begins.

Source: SAP and NVIDIA Open­Shell: Work­ing Toward Gov­er­nance and Secu­ri­ty for Auditable AI Agents in Enter­prise Systems

Curi­ous about what AI agents could mean for your SAP envi­ron­ment? Let’s explore where AI could cre­ate real val­ue in your busi­ness. Get in touch: