Process Intelligence: The Missing Foundation for Administrative Digitization and AI

Executive Summary

Digital transformation in German public administration fails less due to a lack of technology than because administrative processes are insufficiently understood, documented, and managed. Process intelligence—understood as the public sector’s ability to systematically capture its processes, analyze them in a data-driven way, steer them in a targeted manner, and continuously improve them—forms the operational foundation of effective digitalization and responsible AI use.

Conceptually, process intelligence builds on established Business Process Management (BPM) but extends it through data-driven analysis and AI-enabled decision mechanisms Dumas et al., 2018. From a design science perspective, process intelligence can be understood as a socio-technical artifact that integrates organizational structures, methods (e.g., BPMN, DMN, process mining), and technical systems in order to address the problem of fragmented digitalization in public administration Hevner et al., 2004.

1. The Challenge of Digitalizing Public Administration

1.1 Digitalization without Process Reform

Digitalization initiatives in the public sector often focus on the technical implementation of existing procedures. In practice, this means that analog inefficiencies—such as redundant checks, media discontinuities, or unclear responsibilities—are transferred unchanged into digital systems. The result is digitized legacy processes instead of structural improvements, as documented in the e-government literature on fragmented modernization approaches Janssen & Cresswell, 2006.

Effective digitalization therefore requires prior analysis and redesign of the underlying processes, as formalized in the BPM lifecycle phases of process analysis and process redesign Dumas et al., 2018.

1.2 Structural Root Causes

• Fragmented IT landscapes: A lack of interoperability prevents end-to-end process logic and hinders an end-to-end view of public services Janssen & Cresswell, 2006.
• Low standardization: Similar services are implemented differently across organizations; this undermines the scaling and re-use of digital solutions Scholta et al., 2019.
• Responsibility-oriented structures: Processes are designed along organizational units rather than end-to-end process chains, which public management research describes as a major barrier to process reform Bouckaert & Halligan, 2008.
• Front-end focus: Digitalization often targets citizen-facing touchpoints rather than internal core processes, thereby limiting potential efficiency and quality gains Janssen & Cresswell, 2006.

1.3 AI as an Amplifier of Structural Deficits

AI systems strongly depend on clearly defined processes and consistent data structures. Studies on algorithmic decision support in public administration show that unclear decision rules and heterogeneous data sets lead to bias, opacity, and limited scalability Veale & Brass, 2019. In environments characterized by high process variance, unclear decision rules, and low data quality, AI solutions reinforce existing inefficiencies instead of compensating for them.

2. The Concept of Process Intelligence

2.1 Definition and Positioning in the IS Discourse

Process intelligence describes an organization’s ability to understand, manage, and adaptively further develop its business processes on a data-driven basis. It thus stands in the tradition of BPM, which the IS and management literature describes as a holistic approach to identifying, modeling, analyzing, improving, and automating processes Dumas et al., 2018. At the same time, process intelligence addresses the “missing link” problem between information systems and actual process execution that is emphasized in the process mining discourse van der Aalst, 2016.

In the sense of Design Science Research, process intelligence can be conceptualized as an integrated artifact that brings together methods (BPMN, DMN, process mining), organizational roles (process owners, governance bodies), and technical infrastructure (process-capable line-of-business systems, event logs) in order to address a clearly defined relevance problem—the stagnation of digitalization in public administration Hevner et al., 2004.

2.2 Levels of Process Intelligence

• Transparency: Systematic capture and modeling of as-is processes, typically using BPMN 2.0 and process screenings.
• Analysis: Data-driven assessment of efficiency, quality, and variants, for example using KPI systems and process mining techniques van der Aalst, 2016.
• Steering: Continuous optimization, automation, and integration of AI applications, supported by explicit decision models (DMN) and governance structures.

3. Methodological Foundation: Artifact Building Blocks

From a DSR perspective, process intelligence consists of an ensemble of artifact building blocks—conceptual models, methods, and technical components—that jointly enable a process-intelligent public administration Hevner et al., 2004.

3.1 Process Screening

Process screening is the starting point and serves to systematically capture, prioritize, and assess administrative processes. It operationalizes the process identification and documentation phases in the BPM lifecycle and creates a sound basis for subsequent design and evaluation steps Dumas et al., 2018.

3.2 BPMN 2.0 as a Process Language

BPMN 2.0 enables standardized and cross-organizationally comprehensible process modeling and is established in the IS literature as the de facto standard for process modeling Dumas et al., 2018. For public administration, BPMN creates the precondition for integrating business and technical perspectives on administrative workflows and for supporting model-based automation approaches.

3.3 FIM as a Federal Standardization Framework

The Federal Information Management (FIM) framework complements BPMN with public sector–specific structuring of services, data, and processes. It operationalizes the idea of standardization, which the digital government literature highlights as a prerequisite for cross-organizational re-use and scaling Scholta et al., 2019.

3.4 DMN for Modeling Decision Logic

Decision Model and Notation (DMN) enables explicit, formally structured representation of decision rules. From an AI governance perspective, DMN provides a transparent foundation for rule-based and AI-supported decisions and is crucial for the traceability and auditability of algorithmically supported administrative decisions Veale & Brass, 2019.

3.5 Process Mining as Data-Driven Analysis

Process mining closes the gap between modeled to-be processes and real as-is executions by using event data (event logs) to reconstruct actual process variants, bottlenecks, and compliance deviations van der Aalst, 2016. For public administration, process mining provides the empirical basis on which process intelligence can evolve from a purely model-driven approach into a data-driven capability for process steering.

4. Process Intelligence as a Precondition for AI in Public Administration

4.1 Design Relevance: Why AI Fails without Process Intelligence

From the perspective of the DSR relevance criterion, process intelligence addresses a clearly defined practical problem: AI pilot projects in public administration often remain isolated, non-scalable, and difficult to explain. The literature on algorithmic decision-making in the public sector points in particular to poor data quality, unclear decision logic, and weak governance structures as key causes Veale & Brass, 2019.

4.2 Maturity Model as a Conceptual Artifact

The outlined maturity model (from “ad hoc” to “AI-ready”) can be understood as a conceptual artifact in the sense of Hevner et al., 2004. It structures the development paths of public organizations and makes it possible to plan and evaluate design decisions along defined maturity levels (documentation, standardization, data-drivenness, AI integration).

• Ad hoc: Undocumented processes, no consistent view of workflows.
• Documented: Modeled processes, but not firmly institutionalized.
• Standardized: Binding process models, clear roles and governance.
• Data-driven: Continuous measurement, monitoring, and improvement.
• AI-ready: Integration of AI systems into stable, standardized, and data-driven process landscapes.

5. Impact Potential and Evaluation

5.1 Efficiency, Quality, and User Orientation

Empirical studies on BPM show that structured process design and management lead to measurable gains in efficiency and quality, for example in the form of shorter throughput times, lower error rates, and improved service quality Dumas et al., 2018. In a DSR setting, these metrics can serve as evaluation criteria for process intelligence artifacts.

5.2 Transparency, Legal Certainty, and AI Governance

Explicit process and decision models increase traceability, auditability, and legal certainty—a core requirement in the public sector. In the debate on “algorithmic accountability,” it is emphasized that transparent decision rules and documented workflows are prerequisites for legitimate AI-supported decisions Veale & Brass, 2019.

5.3 Scalability and Re-use

Standardized models following FIM and BPMN logic support the “one-for-all” principle, which the German-speaking e-government discourse views as key to scaling digital public services Scholta et al., 2019. From a DSR perspective, this represents an important criterion for the broad impact of the artifact.

5.4 Adaptability and Resilience

Documented and standardized processes facilitate the implementation of regulatory changes and organizational adjustments. The public administration and public management literature discusses this as a central dimension of administrative resilience and change capability Bouckaert & Halligan, 2008.

6. Implications for Design Science Research

6.1 The Artifact Nature of Process Intelligence

Process intelligence can be understood as a comprehensive socio-technical artifact in the sense of Hevner et al., 2004. It combines conceptual models (maturity model, levels of process intelligence), methods (screening, BPMN, DMN, process mining), and technical implementations (event-based system logs, workflow systems) and thus addresses a key relevance problem of administrative digitalization.

6.2 Research Design and Evaluation Paths

For a DSR paper, formative evaluations along the maturity levels and summative evaluations using metrics for efficiency, quality, transparency, and scalability are particularly suitable Hevner et al., 2004. In addition, applying the Hevner guidelines to process intelligence artifacts themselves can become the subject of a conceptual or empirical study Gregor & Hevner, 2013.

7. Conclusion

Process intelligence is a central precondition for successful digitalization of public administration and scalable AI deployment. In the IS discourse, it links BPM, process mining, and AI governance into an integrated, design-oriented approach that addresses the relevance problem of fragmented and technology-driven digitalization initiatives Dumas et al., 2018van der Aalst, 2016. As an artifact in the sense of design science, process intelligence offers a structured framework for integrating and systematically evaluating process, data, and AI perspectives in public administration Hevner et al., 2004.

Literature

• Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004). Design science in information systems research. MIS Quarterly, 28(1), 75–105.
• Gregor, S., & Hevner, A. R. (2013). Positioning and presenting design science research for maximum impact. MIS Quarterly, 37(2), 337–355.
• Dumas, M., La Rosa, M., Mendling, J., & Reijers, H. A. (2018). Fundamentals of business process management (2nd ed.). Springer. https://doi.org/10.1007/978-3-662-56509-4
• van der Aalst, W. (2016). Process mining: Data science in action (2nd ed.). Springer.
• Janssen, M., & Cresswell, A. M. (2006). Enterprise architecture integration in e-government. Proceedings of the 39th Annual Hawaii International Conference on System Sciences. IEEE.
• Scholta, H., Mertens, W., Kowalkiewicz, M., & Becker, J. (2019). From one-size-fits-all to user-centered digital public services: A literature review and research agenda. Government Information Quarterly, 36(1), 11–25.
• Bouckaert, G., & Halligan, J. (2008). Managing performance: International comparisons. Routledge.
• Veale, M., & Brass, I. (2019). Administration by algorithm? Public management meets public sector machine learning. Public Management Review, 21(4), 615–642.