The Exception Explained in Words
SAP runs core finance, procurement, supply chain and operations for large enterprises. The documents that explain any given transaction are usually somewhere else, which is how a variance code becomes the only available answer to a question a person could answer in a sentence. AI Intelligence Hub reads the record and the paperwork together.
How it connects
SAP Integration Suite publishes an MCP server as a first-class artifact, generated from a managed API, an HTTP endpoint or an RFC destination. That shape is worth understanding, because it is better than a fixed endpoint rather than worse: the tools an agent sees are the ones the organization chose to expose, so a finance team can open purchase orders and goods receipts without opening payroll. The platform connects to MCP servers as a client, configured once with a name, the server URL, the transport and the headers carrying its credential, and the tool list is fetched when the connection binds rather than coded in.
Two things follow that an engagement should establish rather than assume. The Integration Suite artifact needs a service plan that carries it, Enhanced or above, so confirm what the organization holds. And SAP publishes no ready-made server for S/4HANA business data. The ABAP and HANA Cloud servers that are generally available are a developer surface and a data-layer surface, not the business-record surface this describes. What is used here is the server the organization stands up over its own APIs.
Where a step needs something the server does not expose, the HTTP Request node calls the API directly, and the platform's own REST API and a workflow's HTTP trigger are the inbound paths, so automation on the SAP side can start a workflow.
The credential the connection carries is the boundary on the SAP side, so what an agent reaches is what that credential is granted. Where a step writes back or sends something, a human-in-the-loop node puts a person in front of it first. Pulling attachments into the library on a schedule is a separate piece of work, built per engagement rather than prebuilt.
What you can do together
- Match an invoice to its purchase order and goods receipt, with the documents behind each read together rather than in three systems.
- Get an exception explained in words with the sources cited, instead of a variance code somebody has to interpret.
- Expose only the APIs the use needs, since the MCP server is generated from what the organization chooses rather than from a fixed catalogue.
- Read an inspection report beside the transaction it belongs to.
A scenario
- ScopingThe engagement confirms the Integration Suite plan, and which managed APIs the MCP server is generated from.
- SetupThe server is created in Integration Suite over those APIs and added once to AI Intelligence Hub with a scoped credential.
- OvernightA workflow reconciles the day's invoices against orders and receipts and lists what does not match.
- A morningFor one invoice the answer explains that the delivery was short, citing the goods receipt and the delivery photograph.
- ApprovalA drafted credit request waits at a human-in-the-loop node until a clerk accepts it.
What stays where
SAP remains the system of record
Transactions, the ledger and the approval structure stay there.
The exposed surface is your choice
The Integration Suite server is generated from the APIs you nominate, so what an agent can reach is decided on your side, not by a vendor catalogue.
No ready-made business-data server exists
The generally available ABAP and HANA Cloud servers are a developer and a data-layer surface; this path uses the one you stand up.
Writes wait for a person
A human-in-the-loop node comes before anything written back or sent.
Products and solutions
Next step
See it on your own SAP instance. We will show the connection made, the data moving and the output, then size it for your deployment.
Request a demonstration or write to sales@vidizmo.ai
sales@vidizmo.ai · vidizmo.ai/integrations/catalog/sap