We deploy SAP Joule assistants and agents, build custom agents for your own processes, and connect SAP to the enterprise AI you already use through secure interfaces, including the Model Context Protocol (MCP). Human approval is built in where it matters, and security, access and audit trail are designed from day one.
SAP Joule
deployed and configured for your SAP landscape
Custom agents
built for your own processes and rules
Your choice of AI
Joule, or an enterprise model connected through MCP
Safe by design
permissions, approvals and a full audit trail
The question is no longer whether AI can work in SAP. It is which work to give it, how much it may do on its own, and how you stay in control.
What we do for you
Choose an area to see the problems we solve, what we deliver and the technology we use.
Deploy SAP Joule
The problem
- Joule licensed but hardly used
- Users unsure what Joule can do for them
- No plan for which scenarios to switch on
What we deliver
- Joule set up across SAP Cloud ERP, S/4HANA and SuccessFactors
- Scenarios chosen and switched on by role
- Identity, roles and data access configured
- Training and adoption so people actually use it
- SAP Joule
- SAP Business AI
- SAP Build Work Zone
The result: People get answers and actions from SAP in plain language.
Build custom agents
The problem
- Standard scenarios don’t cover your processes
- Routine work still done by hand every day
- Ideas stuck at the demo stage
What we deliver
- Agents for your own processes, such as invoice matching, order status, period-end close tasks and supplier follow-up
- Agents built in Joule Studio or on SAP BTP
- Connections to SAP and non-SAP systems
- Pilots in our Co-Innovation Lab, then production roll-out
- Joule Studio
- SAP BTP
- SAP AI Core
The result: Agents that take real routine work off your teams.
Bring your own AI model
The problem
- Joule cost doesn’t fit every use case
- The company already uses another enterprise AI model
- Worry about being locked in to one AI vendor
What we deliver
- Secure connection between your AI model and SAP through the Model Context Protocol (MCP)
- Tools that expose only approved SAP data and actions
- The same permissions and audit as SAP itself
- A mix of Joule and your own model where each fits best
- Model Context Protocol (MCP)
- SAP APIs
- SAP Integration Suite
The result: AI choice and cost control, without losing security.
Guardrails and governance
The problem
- Unclear who is responsible for what an agent does
- Concern about data leaving the company
- No audit of AI decisions
What we deliver
- Role-based access for every agent
- Human approval for actions that matter
- Full logging of questions, answers and actions
- Review of AI provider data terms before go-live
- SAP Cloud Identity Services
- SAP GRC
- Audit logging
The result: AI the business, IT and auditors are comfortable with.
Run and improve
The problem
- Agents that drift and give worse answers over time
- AI costs rising without anyone watching
- SAP releases that break agent connections
What we deliver
- Monitoring of accuracy, usage and cost
- Regular tuning of prompts, tools and rules
- Testing with every SAP release
- New agents added from your use-case list
- iCare AMS
- SAP Cloud ALM
- Monitoring dashboards
The result: Agents that keep working, and keep paying back.
From idea to agent in production
Assess
Rank AI use cases by value and effort.
You get: A ranked use-case list
Design
Choose Joule, a custom agent or your own model, and set guardrails.
You get: An agent design
Pilot
Build and test with your users in the Co-Innovation Lab.
You get: Measured results
Deploy
Roll out with roles, approvals and monitoring.
You get: Agents in daily use
Run
Monitor, tune and add the next agents.
You get: Agents that keep working
Joule or a custom agent?
SAP Joule fits when
- The scenario is already delivered by SAP
- Users work mainly inside SAP applications
- You want SAP to maintain and update the capability
A custom agent or your own model fits when
- Your process is specific to your business
- The work spans SAP and non-SAP systems
- You need to control AI cost or use a model you already have
Related: AI Agents for SAP (agent catalogue) · Autonomous Enterprise Roadmap · SAP Business AI Platform & BTP · Hire AI engineers
Find your first agent
Tell us which process takes your team the most time. We’ll show you whether Joule, a custom agent or your own model fits best.
Frequently asked questions
What is SAP Joule?
SAP's AI assistant, built into SAP applications. It answers questions and carries out tasks in plain language, using your SAP data and permissions.
Do we have to use Joule?
No. Joule fits many SAP scenarios. Where cost or flexibility matters, we can connect an enterprise AI model you already use to SAP through MCP, or combine both.
What is MCP?
The Model Context Protocol, an open standard that lets AI models use approved tools and data safely. We use it to give AI agents controlled access to SAP.
Can an agent change data in SAP?
Only if you allow it. Agents can answer only, draft actions for approval, or complete routine cases within rules you set, always with an audit trail.
Where do we see example agents?
See AI Agents for SAP for the agents we build, and pick the ones you would like to pilot.