MiniCamps

We explore how organizations can move beyond model-centric governance toward decision-centric governance. Using decision-centric orchestration and BPM+ concepts, we examine a practical framework for classifying AI participation across a spectrum that includes AI tools, assistants, performers, agents, and AI orchestrators. We also explore how governance requirements evolve as autonomy increases and introduce a Human-AI Agency Scale (HAAS) for reasoning about responsibility, oversight, and decision authority.

Rule Engines are very powerful tools to encode custom declarative logic — in this talk we look at what benefits we can get from applying Drools to the Kubernetes Control Plane. From symbolic AI and inference to CEP, this might be a good fit to solve common Kubernetes administration problems. This led to “RuleOps,” an exploration space providing a toolkit for cloud engineers and DevOps.

There is so much hype around statistical AI. But why? These AI systems can only reproduce old facts, produce unreliable outcomes, and are not transparent, with architectures that seem to increasingly factor out human control. So why is the decision-making industry heading in this direction — and why hasn’t symbolic AI, which doesn’t suffer from these disadvantages, turned the tide by offering a more reliable and transparent alternative? Perhaps symbolic AI systems are perceived as too hard to implement because they require explicit domain knowledge. Could building them be made easier — could natural-language specification of domain knowledge be an answer? This talk addresses these questions with a presentation and demo of an experimental Natural Language Execution (NLE) platform called Hapsah.

Jacob shares the recent OpenRules experience building decision services capable of handling huge payloads with sound performance, describing how putting a decision service into a cloud-based environment supporting parallel execution allowed a large US corporation to improve performance 100 times.

Vincent shares his experience applying DMN to build the Digital System for Environment and Planning Act in the Netherlands. Issues discussed: how can we minimize translations and directly execute DMN? How do we combine models from different sources to deliver a single conclusion? How do we implement dynamic questioning, and how does this help end users? How do we deal with the declarative character of DMN in dialogue with the end user? Is DMN suitable for modeling legislation, and can models be easily read by legal experts?

Denis presents a summary of the work done by the DMN 1.4 Revision Task Force (RTF) and opens discussion on what should come next for the Decision Model and Notation standard.

We discuss how to learn constraint models for combinatorial problems from positive and negative example solutions. The constraints learned are either structural, coming from the perceived problem structure of the model, or based on input data which provide parameters and index sets. The learning process uses information about global constraints from the Global Constraint Catalogue, and defines a set of constraint patterns to discover conjunctions of similar constraints over common data structures. The CAT tool generates constraint models in MiniZinc, which can then be run by a variety of back-end solvers; it also produces a human-readable description of the generated constraint model. It was tested on the benchmark problems of the PTHG21 Challenge.

Inspired by the solutions submitted to February’s decision-modeling challenge, this talk looks at aspects of managing large decision tables, comparing the performance of various solutions and suggesting things to consider when working toward an optimal approach for large decision tables.

When migrating to an enterprise BRMS from an existing tool for maintaining business logic, implementers are often given a specification of business rules to work from that replicates design patterns from the legacy tool — which can replicate logical problems and inefficiencies in the new BRMS. Inspired by a demonstration compressing over 16,000 rules into just over 1,000 in the February decision-modeling challenge, this talk discusses how Progress Corticon BRMS helps rule modelers catch these logical inefficiencies during the design phase.

“The Prince” by Machiavelli was meant as an instruction guide for new princes and royals. But what if Machiavelli had the benefit of modern modeling tools to capture his teachings? This session revisits “The Prince” using unambiguous terminology and rules to define and automate decision and process models — armed with an automated set of decision services, any layman can aspire to become a Prince.

Agile has been the buzzword across companies and domains for years, with early delivery of value and “pivoting” becoming even more important in recent times. So how does decision modeling fit with Agile practices? How can a complex decision model be developed iteratively so it realizes value as early as possible? How does a user story and its acceptance criteria relate to a decision model? This presentation tackles these questions in practical terms using real-world examples.

Decisions made in business systems at scale determine not only the financial success of an organization but its very character. Consequently, governance over the design, testing, and deployment of those decisions becomes critical to enterprise success. This talk discusses the principles of governance in Sapiens Decision and the means for its implementation.

Practical examples of how to build cloud-native, event-driven business applications using Kogito, a framework based on Drools, jBPM, and OptaPlanner, integrated with Quarkus.

A demonstration of how to build and run a credit-evaluation process using ProcessMaker + OpenRules decision services connected via REST API calls.