Encoded Workflow Graphs for Ad Hoc Automation Recommendations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing process mining tools are ineffective in handling ad hoc processes and require manual identification of automation opportunities, which is time-consuming and inefficient for multiple teams.
Innovation Solution
An automated ad hoc process recommendation system that generates automation recommendations by analyzing workflow data, creating event logs, learning embeddings of process flow graphs, and ranking their amenability for automation using structural complexity measures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If manual identification of automation opportunities is used, then process mining tools can handle predefined workflows, but it is time-consuming and inefficient for multiple teams
Solution Approach 1:
The system enables automated self-identification of automation opportunities by having the process mining tool automatically analyze event logs, extract process flow graphs, generate embeddings, and rank automation candidates without requiring manual intervention from multiple teams
Solution Approach 2:
The patent replaces the manual mechanical process of identifying automation opportunities with an automated computational system that uses graph extraction, embedding generation, and machine learning-based ranking to automatically identify and prioritize automation candidates
2Adaptability or versatility
If existing process mining tools are used for predefined tasks, then workflow automation can be achieved, but they are ineffective in handling ad hoc processes
Solution Approach 1:
The system dynamically adapts to ad hoc processes by using embedding techniques that capture the structural and attribute features of process flow graphs, allowing the automated recommendation system to handle variable and unstructured ad hoc workflows rather than being limited to predefined task templates
Solution Approach 2:
The patent transforms process flow graphs into embedding representations that encode multiple features (structural complexity, node attributes, edge properties), enabling the system to analyze and adapt to diverse ad hoc processes by changing the parameter representation from fixed predefined tasks to flexible feature-encoded graphs
3Productivity
If manual segregation of insights is performed, then automation opportunities can be identified, but the effort and time required increases significantly
Solution Approach 1:
The patent merges multiple manual steps (process extraction, analysis, automation opportunity identification, and ranking) into a single automated integrated system that processes event logs through graph extraction and embedding generation to simultaneously produce ranked automation recommendations, eliminating the need for separate manual segregation efforts
Data Source
AI summary
Methods, systems, and computer program products for generating automation recommendations for ad hoc processes are provided herein. A computer-implemented method includes obtaining workflow data comprising descriptions associated with one or more dynamic processes; creating event logs based at least in part on the descriptions; applying a graph extraction process to derive process flow graphs from the created event logs; generating embeddings of the process flow graphs, wherein the embeddings encode at least one of: one or more structural features and one or more attribute features of the process flow graphs; and identifying at least one of the process flow graphs to be automated based on the generated embeddings; and outputting the identified at least one process flow graph to at least one of: a user and a robotic process automation tool.