Skip to content
Innovativity

For Finance and IT teams live on S/4HANA or ECC, on-premise included

Your S/4HANA is live. Find out where AI is worth it.

Two steps. First we work out which repetitive Finance tasks are worth automating, which ones SAP's own Joule agents cover, and whether your system can run them. Then one case is built and tested on your own data.

  • Runs in your own tenant
  • Data stays in the EU
  • Classified under the EU AI Act
  • A person approves every posting

What a scan result looks like

  • Bank statement cash applicationSAP already does it
  • Manual accrual proposalsSAP already does it
  • Commentary on the monthly margin varianceWorth building
  • GR/IR account reconciliationSAP already does it
  • New company code set-up in customizingWorth building
  • One-off manual journal entriesLeave it alone
  • Cost center and profit center requestsWorth building

The route, in two steps

Each step ends with something in writing. The scan tells you whether a proof of concept is worth paying for, and which one.

Step 1Fixed fee. Two to three weeks.

Finance AI Opportunity Scan

We find the repetitive work in your Finance team and tell you what to do with each task.

  1. 1

    Inventory

    We walk through close, reconciliation, intercompany, accruals and master data with the people who do the work, and count the hours.

    Deliverable: Task list with hours per month

  2. 2

    Sort

    Every task goes on one of three piles: SAP already does it, worth building, or leave it alone. For the first pile we check each SAP agent against your edition, release, licence and design.

    Deliverable: A verdict per task, and a fit check per SAP agent

  3. 3

    Rank

    The worth-building pile ranked by hours saved per month against build and run cost, data readiness and posting risk. Each case gets a business case, a named owner, its EU AI Act class and a data residency note.

    Deliverable: Ranked shortlist with a business case per case, and a PoC brief

Step 2Only if the scan says it is worth it.

Proof of Concept

Your own partner builds the top case, or we do. Then we test whether it holds up in a real close.

  1. 4

    Build

    A working prototype of the top case, built from our brief by your own partner or by us, inside your own Azure or SAP BTP tenant.

    Deliverable: Working prototype

  2. 5

    Verdict

    We run it against a closed period. Do the postings come out right, is every proposal logged and approved by a person, do authorizations and segregation of duties hold, would it survive month-end.

    Deliverable: Go or stop, in writing

Your auditor will not accept "the AI did it"

An agent that posts in SAP is a user like any other. Before it touches your system, we test it the way we test a design.

A person stays accountable for every posting.

What we test before an agent touches SAP

  • It works with the user's own authorizations. No super user and no shared technical account with wide access.

  • Segregation of duties still holds. The agent cannot prepare and approve the same posting.

  • Approval workflows stay in place. The agent prepares, the approver approves.

  • Every action is logged against a person. An auditor can follow who asked, what the agent did and who released it.

  • There is a manual fallback. When the agent is down the process still runs, and the team knows how.

Still on-premise? You do not have to wait for the cloud.

SAP ties its own AI to a cloud commitment. Its finance agents are released for S/4HANA Cloud Private Edition, and Joule reaches on-premise S/4HANA and ECC only for customers who move most of their landscape to RISE with SAP. An agent built in your own tenant works on the system you run today: S/4HANA on-premise, private cloud or ECC.

  • S/4HANA on-premise
  • S/4HANA private cloud
  • ECC

A use case you can picture

New company code? Let the agent do the customizing.

Take the load off your IT department. No more copying values into configuration tables.

A new company code is a long checklist: the company code and its assignments, number ranges, posting period variants, document splitting, and the master data that goes with it. An agent works through the whole checklist from the approved business input and creates every customizing and master data element in your development system, on a transport your team reviews and releases.

  1. Approved business input
  2. The agent configures
  3. Your team reviews and releases the transport

The same pattern works for plants, sales organisations, cost center and profit center roll-outs.

Your Finance data stays yours, and stays in Europe

Two questions come before any use case: where does the data go, and what does the EU AI Act ask of you. Both are answered in the scan, before anything is built.

Runs in your own tenant

The prototype is deployed in your Azure or SAP BTP environment, under your own identity and access rules. Nothing sits in a tool somebody else hosts.

Data stays in the EU

Models are called in an EU region. Your postings and prompts are not used to train anyone else's model. Both are written into the PoC brief.

Your choice of model

Not tied to one model vendor. Where data may not leave your network at all, the case is built on a model that runs inside it.

A person decides

The prototype prepares, your accountant approves. Every proposal is logged with the source data behind it, so an auditor can follow it.

Where the EU AI Act stands

  • AI literacy for staff who work with AIArticle 4
    Applies now
  • Transparency dutiesArticle 50
    Applies now
  • High-risk systemsArticle 6 and Annex III
    From 2 December 2027

Regulation (EU) 2024/1689 as amended by Regulation (EU) 2026/1744. Status November 2026. We give technical and operational readiness, not legal advice. Confirm the classification with your legal counsel.

What that means for a Finance case

Most Finance automation, such as accrual proposals or reconciliation, falls outside the high-risk list. A case that assesses people, such as the creditworthiness of a private customer, does not.

  • Every shortlisted case classified, with the reasoning in writing
  • A data flow per case: what leaves SAP, where it goes, who can see it
  • An AI literacy session for the Finance team that will use the prototype

Three piles. Most AI pilots skip the first one.

That is how companies end up paying to build what they already own. The sort is the part of the scan that needs someone who knows what S/4HANA does out of the box.

  • SAP already does it

    Joule agents, embedded AI and standard apps that sit in your licence or one step away from it. Switching on beats building.

    You get: what to activate, and what it needs from your release, licence and data.

  • Worth building

    Repetitive, rule-heavy, the data is there, and SAP has no answer for it. These go on the shortlist.

    You get: a ranked list and a brief a builder can start from.

  • Leave it alone

    Too rare, too much judgment, or the real cause is a design gap. Automating a workaround only makes the workaround permanent.

    You get: the reason, and the design fix where there is one.

SAP ships its own finance agents. Can your system run them?

Joule assistants for financial closing, receivables, and cash and treasury are released for S/4HANA Cloud Private Edition. Each one comes with conditions. The scan checks them before you commit budget.

Where SAP has an agent and you can run it, we say switch it on. Where you cannot, or SAP has none, the case goes on the build list.

What we check per SAP agent

  • Edition and release. The assistants are released for private cloud, on releases 2023 and 2025. Each agent has its own minimum feature pack and its own set of notes.

  • Availability and maturity. Several agents are open to early adopter customers only. SAP rates each assistant against its own roadmap: some are complete, others stand at a quarter. We plan on what is delivered.

  • Design fit. An agent builds its proposals from your configuration and your posting history. A design gap is repeated at machine speed.

  • Licence and cost. The agents run on AI Units on top of your subscription. SAP's value case uses a generic company. We redo it with your own hours and volumes.

  • Controls. Who reviews the proposal, who posts it, and what your auditor sees afterwards.

Status November 2026, based on SAP Note 3789370 and the notes it refers to. Availability changes with every feature pack.

Where we look

The work that returns every month and that your team would rather not do by hand.

Period-end close

Recurring journal entries, allocation runs, chasing open close tasks.

Reconciliation

GR/IR clearing, bank statement items, subledger against the General Ledger.

Intercompany

Matching differences in ICMR, recharges, netting and the follow-up mails.

Accruals and provisions

Purchase order accruals, deferrals, reversals and the spreadsheets behind them.

Master data requests

Cost centers, profit centers, G/L accounts, supplier and customer changes.

Reporting and commentary

Variance explanations, margin commentary, the monthly pack.

Repetitive configuration

New company codes, plants and organisational units: customizing and master data created by an agent from approved input.

What changes when you start with the scan

  • Without itA vendor demo decides which use case you pilot.

    With itThe hours your own team spends decide.

  • Without itYou build something SAP already ships.

    With itStandard capability is found first and switched on.

  • Without itThe prototype works in the demo and nobody checks the postings.

    With itIt is tested against a closed period, with a written verdict.

  • Without itYou buy SAP's agents and find out later that your release or design cannot run them.

    With itEach SAP agent is checked against your edition, release, licence and design first.

  • Without itYour Finance data goes into a tool somebody else hosts.

    With itThe build runs in your own tenant. Nothing leaves it.

Who does what

We scope and validate. The build is done by your own partner, or by our build team.

Innovativity

  • The scan: inventory, sort and ranking
  • EU AI Act class and data residency note per case
  • The PoC brief, with the posting logic spelled out
  • Testing the prototype against a closed period
  • The written go or stop advice

The build

  • The prototype, built in your Azure or SAP BTP tenant
  • Security and access set up to your IT standards
  • Done by your own partner, or by our build team

Your team

  • A few hours from the people who do the work
  • A test system and a closed period to test against
  • The decision. An accountant approves every posting the prototype prepares.

Start with the scan.

Thirty minutes on a call is enough to tell whether a scan makes sense for your team. Ask for a quote on the same call.