Many CEOs, CFOs and CIOs in mid-sized companies are asking the same question: how do we digitalise procurement in a way people can actually feel, without replacing the ERP system or launching a lengthy transformation programme? Expectations for efficiency and transparency keep rising. At the same time, budgets, IT capacity and the organisation's patience are limited. Large programmes often take years before they pay off.
We see a more pragmatic route: the core system remains the reliable foundation for purchase orders, supplier master data and invoice posting. On top of it, targeted AI applications take over specific, clearly defined tasks in procurement, precisely where a lot of manual work happens today.
Why the core system should stay
In most companies, the ERP system has grown over many years. It holds the approval rules, posting logic and master data that procurement, finance and auditors rely on. That stability is an asset you should not give up lightly.
The real bottlenecks are rarely in the posting itself. They sit in the work before and around it: reviewing documents, reconciling data, chasing missing information, resolving discrepancies. This is exactly where AI applications come in. They complement the core system rather than replacing its central functions.
For a mid-sized energy company, we ran a strategy assessment covering organisation, processes and systems together. The outcomes included automated purchasing of C-parts, a streamlined ordering process and a roadmap for the S/4HANA transformation. The lesson: quick wins in day-to-day operations and a long-term system strategy are not mutually exclusive.
Two concrete use cases
Supplier onboarding and qualification: An AI application can read submitted documents, check them for completeness and consolidate information from different sources. It flags missing details and prepares the qualification against defined criteria, such as certificates, company data or compliance evidence. Final approval and creation of the supplier still follow your established review and approval processes.
Invoice verification: AI can extract invoice data and match it against the purchase order, goods receipt and agreed terms. It flags discrepancies, missing documents or potential duplicate invoices and prioritises cases for the person responsible. Routine cases are handled faster, while exceptions remain clearly visible for review.
In both cases, the AI works with the company's own rules, documents and workflows. The AI prepares, people decide.
Cutting time and cost where it counts
For clearly defined use cases, the initial implementation time of an AI solution can be up to almost 90 percent shorter. In suitable cases, implementation costs can be up to 80 percent lower than with a comprehensive conventional implementation.
These figures refer to the implementation of individual solutions. They are target and experience values from our practice, not a guarantee. The actual effort depends on factors such as data quality, interfaces, security requirements and process complexity.
Ongoing time savings in procurement should also be measured with concrete KPIs, for example:
- Processing time per supplier onboarding
- Share of complete document submissions
- Share of invoices requiring no manual rework
- Time to resolve a discrepancy
What makes implementation work
Whether an AI application holds up in daily operations rarely depends on technology alone. Decision-makers should clarify three points from the outset.
Data and interfaces: The application needs secure access to the relevant data in the core system, ideally via standard interfaces. Changes to the ERP itself should remain the exception.
Responsibility and approvals: It must be clear which steps the AI prepares and who gives final approval. Existing review and approval processes remain binding.
Team buy-in: Involve buyers early. People who see routine work disappear and exceptions surface more clearly will actually use the solution.
Next steps: start small and focused
A mid-sized company can begin with one narrowly defined procurement process:
- Define the core system and approvals: Which system owns the data, and which approval steps are mandatory?
- Map the manual work: Where do recurring tasks and frequent errors occur?
- Pick one use case: for example, the completeness check during supplier onboarding.
- Connect securely and pilot: Link the AI application to data and systems and test it with real cases.
- Measure, then scale: Assess time, cost, quality and acceptance before extending the approach to other processes.
Transforming procurement does not have to start with replacing the ERP system. The core system secures the central transactions, while targeted AI applications handle complementary tasks faster and with a better fit. The real benefit shows in how much manual work actually disappears from each process.
Want to know which procurement process is the right starting point for you? Our procurement assessment shows where the greatest potential lies. Or book an initial consultation directly.