Perspective · 6 min read
Enterprise AI: the use cases that create real value on SAP.
Between spectacular demos and processes that run overnight, there is a gap. We spend a good part of our time helping clients cross it. Here is how we sort, and why so many "AI everywhere" projects never get past the pilot stage.
Signed by the Casablanca AI Center of Excellence · a clear position, lessons from experience and recommendations
Rarely has a technology been adopted so fast in conversation and so slowly in actual processes. Every leadership team has seen impressive demos. Many have launched pilots. Very few can point to an end-to-end process whose performance has changed for good. Contrary to received wisdom, this gap owes little to the maturity of the models; it comes from the way companies choose their use cases. Or, more precisely, from the way they fail to choose them.
Why "AI everywhere" fails
Sprinkling AI over every subject at once invariably produces the same result: a collection of orphaned pilots. Post-mortem after post-mortem, we keep finding the same mechanisms.
- The demo replaces the process. An assistant that shines in a steering committee and a piece of processing that holds its promise across thousands of real transactions, with their exceptions, their disputes and their imperfect data, are two different things. The first takes a few weeks. The second requires reworking the process around it.
- No one owns the outcome. When the use case is born in a "lab" rather than with a process owner, no one answers for the gap between promise and production. The pilot succeeds, the rollout is orphaned, the subject fades away.
- The data is discovered too late. Enterprise AI does not reason over the internet, but over your orders, your invoices, your master data. If they are incomplete or contradictory, the model industrializes the confusion. Many AI projects die of a data problem that honest scoping would have exposed on day one.
What sets a good use case apart
Our criteria are deliberately austere. Five conditions, all of them necessary.
- A frequent process that is costly in human attention. AI’s value is measured in volume; beware the prestigious, low-volume subject.
- An outcome measurable in the business’s own terms: lead time, error rate, workload avoided. Defined before the project. Indicators rebuilt after the fact prove nothing.
- Accessible, sufficiently reliable data, or a data-quality workstream explicitly included in the scope.
- A business owner who wants the outcome badly enough to accept changing their own process to get it.
- A path to production: system integration, exception handling, supervision. Without an industrialization plan, a use case remains a prototype, whatever its budget.
Replace "where can we put AI?" with "which process do we want to see run differently a year from now?" and half the sorting is already done.
On SAP, what makes it into production
Within the SAP perimeter, the use cases that actually reach production look alike. In practice, we see them cluster into three areas.
Document automation
Supplier invoices, customer orders, confirmations, remittance advices: a considerable share of administrative work consists of reading documents in order to rekey their content into SAP. Assisted extraction, matching and posting are the surest ground for enterprise AI. Volumes are high, the result can be verified, and the gains are visible to the teams themselves, which changes everything for adoption.
Predictive, anchored in operations
Demand forecasting, anticipating supplier delays, cash projection: prediction creates value when its output feeds a recurring decision that already has a process behind it, a supply plan or a risk review, rather than yet another dashboard. The decisive criterion has nothing to do with how sophisticated the model is. Someone in the business has to do something differently because of it.
Copilots and agents
Assistants that answer with the company’s knowledge, agents that prepare or execute transactions under human control: the most promising family is also, in our view, the most demanding. It demands a strict framework. A delimited scope of action, traceability of every step, human validation for anything that commits the company. An agent without governance is just an operational risk with a good interface.
The prerequisites enthusiasm cannot replace
None of these families holds without two foundations. Data, first: clean master data, shared definitions, governed access. It is the thankless workstream on which all the others depend, and the reason our SAP Analytics offering treats data and its uses together. Governance, next: who decides what AI is allowed to do, who supervises its outputs, who answers for its errors. We refuse to see these as brakes on innovation. They are what separates an enterprise capability from a permanent experiment.
And because these foundations live in the core of the system, the AI trajectory cannot be planned separately from the ERP trajectory. A clean S/4HANA core, with harmonized processes and reliable data, remains in our view the best AI investment most companies can make. We explain why in our reading of the ECC 2027 deadline.
One process at a time
Which process do you want to transform?
Bring a process, its volumes and its pain points: thirty minutes with a KOORTIS expert to assess whether AI belongs there, and how to prove it.
Let’s talk →