SAP today added a raft of Joule artificial intelligence (AI) agents across its application portfolio that lays the foundation for building a broader ecosystem.
Announced at the SAP Sapphire conference, SAP also revealed it has developed an application development platform, dubbed SAP AI Foundation, for building AI applications and extended the portfolio of AI applications it provides via the SAP Business Data Cloud to include a People Intelligence offering for managing human talent using data collected from SAP SuccessFactors human resources platform.
Additionally, SAP unveiled a tool that promises to make it easier to modernize existing SAP applications, using its SAP Signavio business process re-engineering platform and SAP LeanIX framework for building user interfaces.
Based on the large language models (LLMs) that SAP used to develop its Joule Copilot tool, the autonomous AI agents developed by SAP are capable of studying end user behavior to provide a more proactive approach to improving productivity by as much as 30%, says Manoj Swaminathan, general manager and chief product officer for the finance and spend applications within the SAP Business suite. “We’re launching a network of autonomous decision-making agents,” he adds.
SAP trained its Joule AI agents using data it gained with the acquisition of WalkMe, a provider of a platform that provides guidance and real-time support directly within applications, in 2024. That approach provides a more omnipresent set of AI agents capable of engaging in more personalized conversations with end users, says Swaminathan.
As part of that effort, SAP has also partnered with Perplexity to gain access to both structured and unstructured data, such as charts and graphs that can be exposed to SAP Joule agents in real-time.
The overall goal is to provide end users with a set of AI agents that better understand the context in which they are being assigned a task to complete, says Swaminathan.
It’s not clear how quickly organizations are adopting AI agents but a Futurum Research report predicts they will drive $6 trillion of economic value by 2028. The challenge now is determining how best to employ AI agents that leverage LLMs to provide probabilistic insights within the context of business applications that are often deterministic in the sense that tasks need to be completed the same way every time. While advances continue to be made in terms of training AI agents using more domain-specific data, there is still a need for humans to review the output created by an AI agent.
Ultimately, AI agents from SAP will need to interact with a multitude of other AI agents that are being developed by both IT vendors and individual IT teams. SAP clearly hopes to provide one of the major frameworks through which those AI agents will be orchestrated. Ultimately, that orchestration framework will then need to be integrated with other frameworks to enable organizations to further automate end-to-end processes.
In the meantime, organizations would be well advised to start identifying which tasks being performed today by humans might soon lend themselves to being performed by AI agents that those humans will soon supervise.
