SAP is adding 15 Joule AI agents to its business portfolio. During SAP Connect in Las Vegas, the company announced new Joule agents. Soon, you will have an agent available for every branch of the business. From cash management to production planning, from HR questions to account planning. SAP takes Agentic AI seriously, but will organizations still be able to keep track of all those agents?
We already knew that SAP was busy with AI and Joule. However, the focus was initially mainly on their ERP solution S4HANA. During SAP Connect, ERP will be discussed, but the primary focus will be on SAP’s other business applications. Think of SuccessFactors (HR), SAP Ariba and SAP Concur (procurement and spending), CX (CRM, Sales and Service), and the supply chain. All these solutions will now have their own Joule agents.
What are Joule agents?
Before we dive into the details: Joule is SAP’s AI co-pilot, a chat window that allows users to communicate with SAP software in natural language. Joule agents are specialized AI assistants that not only provide answers but also perform actual actions. They analyze data, make recommendations, and perform tasks that were previously done manually by the user.
The difference with a traditional copilot is that an agent can independently reason about complex information and then take action without you having to intervene constantly. An example: there is now a Joule agent that takes work off the hands of a financial employee. The employee no longer has to manually reconcile bank statements on a daily basis; the Cash Management agent now does this automatically. Only the rules that deviate remain for a human to assess.
Every complex business function gets a Joule AI agent
We have attempted to provide an overview of all the AI agents that SAP is announcing this week; however, we cannot rule out the possibility that we may have overlooked one or two.
Finance:
The Cash Management agent automates daily reconciliation tasks, according to SAP, saving around 80 percent of the time. The International Trade Classification agent can classify goods for international shipment, providing relevant information for Customs purposes. This is particularly interesting for companies that export a large volume and want everything to run smoothly at Customs.
That 80 percent time saving sounds great, but those are SAP’s own figures. There are certainly benefits to be realized, but in practice, it depends heavily on how an organization works and what the data looks like. The principle ‘garbage in, garbage out’ certainly applies to AI agents as well.
Procurement and purchasing:
The Bid Analysis agent can analyze quotes from suppliers and consider various factors. These include total costs, costs per item, shipping costs, and payment terms. According to SAP, many companies often overlook the details and focus solely on total costs, even though there can sometimes be significant differences and advantages to be gained.
Additionally, a Receipt Analysis agent will be available, automatically filling in missing information on receipts. It does this by using supplier databases, web searches and maps.
Supply chain:
There will also be a specific Joule agent for the supply chain. The Production Planning agent can automatically check whether production orders can be released. This is based on the availability of materials, capacity, and planning. All data is validated before an order is released. Additionally, there is a Change Record Management agent that assists engineers in managing change requests and initiating processes.
Finally, a Supplier Onboarding agent will ensure that new suppliers are registered in the system more quickly and efficiently. We have found, through experience, that this process can be significantly streamlined at many companies.
HR:
With a solution such as SuccessFactors, HR agents cannot be left behind. The first is the HR Service agent, which several organizations have already introduced on their own. This agent answers questions from employees about HR company policy. As a result, HR departments no longer have to answer all simple questions themselves.
There will also be a Career and Talent Development agent who will support managers in discussing their subordinates’ careers and development with them in a timely manner. The agent will also keep the manager informed of the progress made.
The People Intelligence agent will analyze personnel data to identify trends. It can then recommend actions in the areas of retention and compensation.
Sales and customer service:
The Account Planning agent creates strategic account plans by analyzing customer history, purchasing patterns, and business context. This should ensure that sales staff are always well prepared for sales calls without having to spend hours on manual preparation.
The Digital Service agent will offer 24/7 automated customer service in multiple languages and will only escalate to human employees when really necessary. It will attempt to resolve as many tickets and questions as possible on its own.
Agent for utility companies
SAP has also announced a Utilities Customer Self-Service agent for utility companies. This agent will help customers with questions about contracts, rates, and consumption data. It is not yet entirely clear to us whether this is suitable only for specific countries or if this agent can be used worldwide. However, it could be an interesting case, given that the energy market has been somewhat less stable in recent years. Energy companies are undoubtedly receiving more questions in their customer service departments.
Agents who talk to agents
All these agents will not only communicate with people, but also with each other. SAP is firmly committed to Agentic AI. Joule acts as an orchestration layer, a kind of traffic controller that calls the right agent at the right time. But soon, an agent will also be able to decide for itself to consult another AI agent.
This will depend on the context, the user’s role, and ultimately, the agent’s role. To enable these agents to communicate effectively with each other, SAP is working on incorporating the Agent-to-Agent (A2A) protocol. This will not only enable SAP’s Joule agents to communicate with each other, but also with third-party agents. Alternatively, third-party agents can communicate with Joule AI agents. A2A is a standard protocol that has been widely adopted by various IT parties to enable agents to communicate with one another. In addition to A2A, there is also MCP (Model Context Protocol), which SAP also supports.
Managing agents from the AI Agent Hub
An AI Agent Hub will be available within SAP LeanIX. This will be the central location where organizations can manage their AI agents. The goal is to be able to manage not only SAP’s AI agents, but also those of third parties. This is similar to what Workday recently announced with its Agent System of Records or ServiceNow with its AI Control Tower.
All major platform companies want to control and orchestrate AI agents. SAP, Workday, ServiceNow, Salesforce. The problem, however, is that they can only manage their own agents, not those of third parties, because there is no standard protocol for this yet. These are ideas and concepts, but the market is not there yet.
The future of enterprise software is AI agents, but is this the best way?
On the one hand, it makes sense that SAP is presenting at least 15 Joule agents this week. All of these agents will undoubtedly do work that used to take a lot of time, and that really helps employees. This is a trend we see in many SaaS solutions; everyone is adding various AI agents.
However, we are starting to wonder whether users won’t get lost in the availability of these huge amounts of AI agents all serving a specific task. If we continue at this pace, the average business user will soon be interacting with more than 100 agents in a week. That no longer sounds like a tool, but like excess.
Guidance on deploying AI agents
We understand that all these SaaS vendors want to equip their business applications with AI agents, but with numerous separate AI agents and possibly different names, it quickly becomes overwhelming and confusing. Furthermore, companies also need to carefully consider the adoption of AI agents, which requires guidance. Ideally, the system should proactively ask: “I see that you are reconciling manually, would you like the Cash Management agent to take over?” Then people don’t have to familiarize themselves with all those agents in advance. SAP could do that perfectly well with its WalkMe acquisition.
Waiting for availability, practical implementation becomes a challenge
With this rollout of Joule agents, SAP is demonstrating a clear strategy. They are taking agentic AI seriously and are going to integrate it into every business process. The fact that they have now integrated agents into virtually every SAP business application is impressive and very useful for large SAP customers who are fully immersed in the ecosystem.
However, there are still some questions that remain. Most agents are not yet available. They will gradually become available one by one over the coming quarters. Organizations will also need to discover how to use these agents effectively without it becoming confusing or complex.
There is also the issue of governance: who is responsible if an agent makes a mistake? What if the Cash Management agent reconciles a transaction incorrectly, or the Bid Analysis agent makes the wrong recommendation?
This applies not only to SAP, but also to other SaaS vendors. Everyone is in the same race, and no one knows how long it will last or where the finish line is.
SAP is innovating rapidly with agents, which are forthcoming; however, the implementation and execution at customer sites now need to occur. The next 12 to 18 months will be crucial. Then we will see whether SAP’s vision of agentic AI will work in practice, or whether we will drown in a sea of specialized agents all doing their own thing.