Predictive Workflow Analytics Platform with ML Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current workflow management technologies in digital workplaces are limited in providing holistic information and impact analysis for users executing tasks, such as CapEx workflows, as they can only display basic data like expenditure costs, making it difficult for approvers to understand the broader enterprise metrics like revenue, profit margin, sustainability, governance risk, and compliance.
Innovation Solution
A predictive workflow and analytics platform integrates machine learning (ML) models with an analytics user interface (UI) into workflow tasks, allowing for data collection and analysis from various sources to provide predicted outcomes and impacts, enabling low-code/no-code integration and native UI rendering across different devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional workflow management technologies are used, then basic data display is simple and straightforward, but holistic information and impact analysis are insufficient
Solution Approach 1:
The patent introduces ML model services as intermediaries between the workflow management system and data sources. These services process raw data and provide predictive insights (e.g., revenue impact, profit margin analysis) without requiring the workflow system itself to become complex. The metadata file acts as another intermediary, bridging the workflow task definition and the ML inference service.
Solution Approach 2:
The system segments functionality into separate components: workflow task definition, metadata configuration, ML model services, and UI rendering. Each component has a specific responsibility, allowing the system to provide comprehensive information without monolithic complexity. The metadata file segments the configuration data from the execution logic.
2Reliability
If comprehensive data analysis is provided, then decision-making quality improves, but data retrieval time increases
Solution Approach 1:
ML models perform predictive analysis in advance, calculating potential impacts (revenue, profit margin, sustainability metrics) before the approval decision is made. This preliminary action provides comprehensive data ready for immediate display, eliminating the need for time-consuming real-time analysis during the decision process.
Solution Approach 2:
The patent replaces manual data retrieval and analysis with automated ML-based predictive analytics. Instead of users manually gathering data from multiple sources, the system automatically substitutes this mechanical process with intelligent models that provide instant, comprehensive analysis.
3Adaptability or versatility
If ML model integration is implemented, then predictive analytics capability is enhanced, but implementation complexity increases
Solution Approach 1:
The patent uses lightweight metadata files instead of complex integration code. These metadata configurations are simple to create and modify, allowing non-expert developers to implement predictive analytics without dealing with complex ML model integration. The metadata acts as a disposable configuration layer that simplifies implementation.
Solution Approach 2:
The workflow task definition and metadata structure are designed to be universal, supporting multiple ML model types and predictive analytics scenarios through a single framework. This multi-functionality allows the same mechanism to provide various predictive capabilities (revenue impact, risk assessment, compliance analysis) without requiring separate implementation approaches.
4Ease of operation
If native UI rendering across devices is achieved, then user experience consistency improves, but development complexity increases
Solution Approach 1:
The metadata-driven UI definition enables a single workflow task definition to render natively across multiple devices and platforms. The same metadata configuration provides consistent user experience on web, mobile, and desktop without requiring separate development code for each platform, achieving universality in UI rendering.
Data Source
AI summary
Methods, systems, and computer-readable storage media for extracting, by a multi-experience runtime engine and from a metadata file, metadata that is descriptive of an analytics UI for display on a display of a computing device, the metadata including instructions for a binding to a service providing inference using one or more ML models, in response to the binding, transmitting an inference request to the service through a predictive data adapter, the inference request including data representative of a workflow task that is to be executed in a digital workplace, receiving inference results that are responsive to the inference request, and displaying, within the analytics UI, the inference results and at least a portion of the data representative of the workflow task.