Häufig gestellte Fragen
What does artificial intelligence in ERP actually deliver?
AI in ERP primarily improves forecasts such as demand and sales predictions, automates recurring routine work such as document recognition and account assignment, and provides conversational assistants that answer questions and execute actions. Added to this is anomaly detection, which surfaces conspicuous postings, potential fraud patterns or quality deviations early. The benefits are real, but they depend heavily on clean master data and well-designed processes and rarely materialise on their own. What is meaningful is therefore not the buzzword but the concrete, measurable use case in your own process.
What specific AI use cases exist in ERP?
Typical use cases are demand forecasting for AI-supported demand predictions, predictive maintenance for scheduling maintenance before a failure occurs, and the automated processing of incoming invoices from data capture through verification to posting. In addition, there is process mining, which uncovers inefficiencies in workflows, as well as chatbots serving as internal self-service assistants. In HR, AI functions are also used for applicant pre-screening and in finance for credit checks, which are legally particularly sensitive. Which of these cases pay off is determined by a high repetition rate, a clearly measurable outcome and available data.
Does AI in ERP replace human decisions?
No, AI supports and accelerates workflows, but for postings and business-critical decisions it does not replace human oversight. This is not only a matter of diligence but also of compliance and liability, because the company using the system legally counts as the operator of the AI system. For applications classified as high-risk, such as applicant screening or creditworthiness checks, the EU AI Act explicitly mandates human oversight, although the obligations for such high-risk systems are, as things currently stand, expected to apply only from the end of 2027. AI is thus a tool that delivers suggestions and preparation, while responsibility for the decision remains with humans.
Which vendors have integrated AI into their ERP?
The major vendors include SAP with its assistant Joule in S/4HANA and the Business Technology Platform, and Microsoft with Copilot in Dynamics 365, closely interlinked with the Power Platform. Oracle also offers numerous embedded AI agents in its Fusion Cloud, while Infor and NetSuite bring their own AI and analytics functions, for example for demand forecasting. Increasingly, cloud ERPs from the DACH region are also expanding AI features for document processing and forecasting. In the selection process, the marketing promise should count for less than the demonstrable impact in your own specific process.
What prerequisites does AI in ERP need in order to deliver value?
The most important prerequisite is data quality, because AI functions are only as good as the underlying master data, and industry surveys cite poor data quality and data silos as the most common cause of failed projects. A step-by-step approach has proven effective, consisting of use-case selection, a data readiness assessment, a pilot with defined KPIs and subsequent scaling. Equally decisive is involving and training the workforce early, because without competence building, measurable results usually fail to materialise. Realistic payback periods are around 6 to 18 months, provided use cases are prioritised by impact, feasibility and data availability.
What obligations apply to AI use in ERP under the EU AI Act?
Anyone using AI in an ERP legally counts as an operator and bears their own obligations under the EU AI Act. Since 2 February 2025, Article 4 has required adequate AI literacy for all employees working with AI systems, regardless of company size and risk class and even where a cloud solution with embedded AI is merely being used. For high-risk systems under Annex III, which include ERP functions such as applicant screening and credit checks, obligations regarding documentation, human oversight and data quality apply; the original deadline of 2 August 2026 is set to be postponed to 2 December 2027 under the politically agreed Digital Omnibus (as of mid-2026 not yet finally adopted). Violations of the high-risk obligations carry fines of up to 15 million euros or 3 percent of global annual turnover, whichever is higher, while prohibited AI practices are punishable by as much as 35 million euros or 7 percent.
What about data protection and the GDPR when using AI in ERP?
AI functions in an ERP regularly process personal data, for instance of employees, applicants or customers, and must therefore be operated in compliance with the GDPR. This requires a sound legal basis, the principle of data minimisation and transparency about which data is used for which purpose. For automated decisions with legal effect, additional specific requirements apply, which is why human control and traceability must be ensured. From 2 August 2026, the EU AI Act additionally requires that AI-generated content and chatbot answers be labelled as such.