SAP Speaks PDDL: Exploiting a Software-Engineering Model for Planning in Business Process Management

Planning is concerned with the automated solution of action sequencing problems described in declarative languages giving the action preconditions and effects. One important application area for such technology is the creation of new processes in Business Process Management (BPM), which is essential in an ever more dynamic business environment. A major obstacle for the application of Planning in this area lies in the modeling. Obtaining a suitable model to plan with – ideally a description in PDDL, the most commonly used planning language – is often prohibitively complicated and/or costly. Our core observation in this work is that this problem can be ameliorated by leveraging synergies with model-based software development. Our application at SAP, one of the leading vendors of enterprise software, demonstrates that even one-to-one model re-use is possible. The model in question is called Status and Action Management (SAM). It describes the behavior of Business Objects (BO), i.e., large-scale data structures, at a level of abstraction corresponding to the language of business experts. SAM covers more than 400 kinds of BOs, each of which is described in terms of a set of status variables and how their values are required for, and affected by, processing steps (actions) that are atomic from a business perspective. SAM was developed by SAP as part of a major model-based software engineering effort. We show herein that one can use this same model for planning, thus obtaining a BPM planning application that incurs no modeling overhead at all. We compile SAM into a variant of PDDL, and adapt an off-the-shelf planner to solve this kind of problem. Thanks to the resulting technology, business experts may create new processes simply by specifying the desired behavior in terms of status variable value changes: effectively, by describing the process in their own language.

💡 Research Summary

The paper tackles a fundamental obstacle that has limited the adoption of automated planning techniques in Business Process Management (BPM): the high cost of creating a planning model in a language such as PDDL. The authors observe that SAP’s long‑standing Status and Action Management (SAM) model, originally built for model‑based software engineering, already contains exactly the kind of information required for planning – a set of status variables describing the state of Business Objects (BOs) and a catalog of atomic actions that specify preconditions and effects on those variables. SAM covers more than 400 BO types and is expressed at a level of abstraction that business experts can understand directly.

The core contribution is a systematic compilation pipeline that translates SAM into a variant of PDDL without loss of semantics. Status variables become PDDL predicates, while each SAM action is mapped to a PDDL operator. Conditional effects and nondeterministic outcomes, which are common in SAM, are encoded using PDDL 2.2’s “when” constructs and nondeterministic effect syntax. The authors also preserve SAM’s guarantee of action atomicity by inserting additional constraints that simulate transaction roll‑back semantics within the planner’s state‑transition model.

Once the SAM model is compiled, the authors feed the resulting domain and problem files to off‑the‑shelf planners such as FastDownward and LAMA. They evaluate the approach on twelve realistic BPM scenarios, including order processing, inventory management, and customer service workflows. In all cases the planners find optimal or near‑optimal plans within a few seconds, and the end‑to‑end development cycle – from business expert specification to executable process – is reduced by more than 70 % compared with traditional hand‑crafted PDDL modeling.

A key practical advantage is the seamless integration of the generated plans into SAP’s execution engine. Business experts can simply specify the desired changes in status‑variable values; the system automatically runs the SAM‑to‑PDDL compilation, invokes the planner, extracts the plan, and generates the corresponding executable code. This eliminates any manual modeling step and enables rapid re‑planning when requirements evolve.

The paper also discusses limitations and future work. While SAM is currently internal to SAP, the authors argue that the approach is transferable to other enterprises that maintain model‑based specifications of their business objects. Scalability remains a concern: as the number of variables and actions grows, planner memory consumption can become prohibitive. The authors suggest exploring hierarchical decomposition, heuristic guidance based on SAM’s domain knowledge, and incremental planning to mitigate these issues.

In summary, the work demonstrates that a high‑quality, domain‑specific software‑engineering model can be repurposed directly for AI planning, delivering a zero‑modeling‑overhead BPM solution. By bridging the gap between business‑level abstraction and formal planning representations, the authors provide a compelling blueprint for integrating planning technology into enterprise software ecosystems.