In the insurance industry and other sectors, digitization gives rise to technical systems that process vast amounts of data driving increasingly complex business processes. The systems also enrich this data using statistics, advanced analytics and machine learning.
There are numerous use cases for linking processes with intelligence (more modestly put: automated analytics). This includes risk calculations, flexible market pricing based on competitor analysis, diagnostic support via decision tables, similarity analyses and customer recommendations in real time, or chatbots that act as a spoken-language interface to complex backend processes.
The decisive competitive advantages gained through this technology are flexibility, adaptability and an accelerated time to market. When business processes or business rules change, certain repetitive tasks and trained machine learning modules can be reused.
The challenge: Infuse business processes with intelligence and generate executable workflows.
- State-of-the art digitization via a robust and expandable Camunda workflow and decision engine
- Faster time to market by directly mapping graphical process notations to executable workflows
- High flexibility thanks to a consistently workflow-driven software design
- Fast integration into existing customer systems via a library of interfaces and additional functionalities (“metaLink”)
- Wide range of options for the tight integration of intelligent components: from rule engines and chatbot dialogs to advanced analytics and machine learning-based decision-making processes
- Customers gain full control of the product and operation thanks to open-source design
- Uniform deployment on different infrastructures, from cloud to blockchain
metaLink is a powerful toolbox that allows you to implement AI-driven process automation. A functionally enriched workflow and decision engine based on the open source platform Camunda makes it possible to efficiently integrate business logic into your existing system components and enrich it with advanced analytics or AI. Design processes graphically in BPMN (Business Process Modeling and Notation) and combine them with DMN-compliant (Decision Model and Notation) decision tables, and use the result to quickly generate executable workflows. You can interact with these via a traditional graphical user interface or using chatbots.
The added value is plain to see: Improved efficiencies through the flexible, adaptive design of complex, microservice-based software systems. And thanks to the open source license, you retain full control over the further development and operation of the systems.
We are continuously expanding the framework to meet evolving customer requirements. Seamlessly integrate it into your existing system landscape, and take advantage of the many applications and options of intelligent automation.
- Kernel extensions: flexible and freely definable data model for the interaction between process steps and separation of the system and application layer for error handling and logging.
- Generic delegates – cross-customer add-on functions and interfaces (APIs): PMML (Predictive Model Markup Language) can be used to integrate advanced analytics and machine learning modules at any point in the process. The Drools rules management system extends the existing DMN engine. This makes it possible to infuse chatbot dialogs with extended intelligence via BPMN tasks. We furthermore implemented database connectors as well as a Salesforce interface and are currently working on integrating this into various blockchain stacks.
- Specific delegates – client-specific enhancements: Integration with identity management or existing backend systems.
Continuous integration is an essential part of modern, agile system development. Our DevOps experts provide you with continuous deployment to various cloud infrastructures.
Business Process Analysis and Modeling
- Analysis of the business processes in collaboration with your analysts
- Modeling in BPMN
- Definition of decision tables
- Optional: Tool-driven automatic process analysis and mapping to BPMN
- Mapping business processes to executable metaLink workflows between system components, taking into account business and technical error handling and logging
- Definition of the data model and the data structures exchanged between workflow steps (tasks)
Development and integration of AI models
- Definition, implementation and, where needed, training of suitable advanced analytics and machine learning models
- Integration in metaLink tasks via PMML
- For more information, click here
Implementation of Tasks (Capabilities) and System Integration
- Implementation of locally executable tasks incl. access to databases
- Implementation of access to external systems or from external systems via the interfaces provided with metaLink
- Design and implementation of communication from and to GUIs
Continuous Deployment, User Acceptance Test, System Integration Test
- Creation of the continuous integration or deployment toolchain
- Deployment and system integration test; testing of the interaction of the various solution components
- User acceptance test (UAT), acceptance test in cooperation with the customer
Our success is our customers’ success.
The key aim of automated diagnostic support is to generate a “Next Best Question” based on the knowledge at hand.
Support of back pain diagnostics
Pilot of knowledge-based decision support for call center employees
The incorrect or unnecessary treatment of unspecific back pain puts a great strain on the public health budget. 80% of all treatment costs for back problems fall into this category. The same issue also often leads to capacity shortages in specialist outpatient clinics.
The aim of the project was to equip the company’s own call center service with an intelligent diagnostic support system. The project objective was to deliver better and more accurate diagnoses via machine deduction based on a powerful knowledge graph.
Simply put, the solution provides possible diagnostic indications in real time after the first symptom has been entered. The result can be refined successively by adding further symptoms. In a later step, physicians evaluate the recommendations and correct the system by providing feedback.
Real-Time Customer Management
High-quality recommendations in real time
Being made aware of new services or products and receiving reminders when visiting a website – these are the kinds of things we are used to from the retail sector. Insurance company websites, on the other hand, are rarely viewed at all. As a consequence, this industry can only gather small amounts of data for analysis. Therefore, generic analytics models are of little use here. Instead, we developed customized advanced analytics models to be used alongside rule sets.
The customer challenge: The company wanted a system that immediately recommends suitable products or services to interested parties who are visiting their website for the first time.
When visitors are offline, our customer’s service can track events from any input sources, group them into sessions and create user profiles. A process-, rule- and model-based recommendation engine then generates recommendations in real time (“Next Best Action” / “Next Best Offer”). A dashboard provides an at-a-glance overview of the reports and helps to perform quality assessments.
metafinanz assisted the customer in the architectural design of the new service and helped implement it – including the development of a machine learning module and the creation of a DevOps infrastructure.
It takes an approach that combines the process model, rule set and machine learning to generate high-quality customer-specific recommendations.
metaLink exemplifies how much process-centric design and flexible automation depend on one another – especially so in productive operation.
Optimal asset strategy for private investors
An asset management company is getting ready to move into the future. It is building a new platform to offer customers a highly automated service. The focus is on private investors.
The automated customer journey begins with a catalog of questions, after which the customer is recommended an optimal investment strategy. This takes into account various parameters such as risk affinity, investment horizon, investment volume and family circumstances. Interested clients are identified via a video identification procedure in keeping with the Money Laundering Act and are provided with a custodial account at the asset manager’s partner bank.
The metaLink workflow engine is the core component driving this process. It contains analytics models as well as other functions and can link up with an internal corporate registration backend. If processes change or additional analytics models are added, adjustments can be made easily without changing existing code.
AIaaS (Artificial Intelligence as a Service)
Camunda Framework PoC
Our customer is planning to set up an AI-driven service platform (AIaaS) to support the flexible implementation of various applications. The solution backend mainly consists of a workflow engine that can be controlled via Business Process Modeling and Notation (BPMN) to orchestrate microservices. The solution is also to include rule-based decision engines, and future user interfaces are required to support chat bots.
First, a provisional application was developed to evaluate the suitability of a Camunda framework with added functions and services with regard to the defined requirements. The results were promising.
“metaLink passes trial run.”
Would you like more information?
Dr. Matthias Besch
metafinanz Informationssysteme GmbH
Leopoldstrasse 146 | 80804 München, Germany
Tel. +49 89 360531-0