AI Training with Accessibility Data for Endpoint Error Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional helpdesk bots integrated with helpdesk services are ineffective in identifying and solving endpoint device errors due to reliance on text-based diagnostics and lack of access to operational data, leading to inefficient manual troubleshooting processes that only address errors after they occur.
Innovation Solution
Training an artificial intelligence model using accessibility data, telemetry, and endpoint agent state information to detect and resolve device errors proactively, enabling automated troubleshooting and preventive measures across multiple devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional helpdesk bots use text-based diagnostics only, then the system complexity is low, but the troubleshooting effectiveness and error detection accuracy are poor
Solution Approach 1:
The patent combines multiple data sources including accessibility data, telemetry data, logs, and state information into a unified training dataset for the AI model. This merging of diverse data types enables comprehensive error detection while maintaining manageable system complexity through integrated processing architecture.
Solution Approach 2:
The AI model is designed to perform multiple functions: detecting errors, diagnosing issues, and recommending resolutions across various endpoint devices and applications. This multi-functional approach improves detection accuracy without proportionally increasing complexity by using a single versatile model rather than multiple specialized systems.
2Productivity
If manual helpdesk processes are used, then the troubleshooting can be performed with simple tools, but the productivity and efficiency of error resolution are low
Solution Approach 1:
The system enables self-service troubleshooting by training the AI model on historical accessibility data, telemetry, and resolution outcomes. The model automatically detects errors and generates resolution recommendations without requiring manual human intervention for each incident, thereby improving productivity while maintaining appropriate automation levels.
Solution Approach 2:
The system implements feedback loops where resolution outcomes are fed back into the training dataset, continuously improving the AI model's accuracy and effectiveness. This automated feedback mechanism enhances error resolution efficiency by learning from past successes while reducing the need for manual troubleshooting processes.
3Reliability
If helpdesk bots react to errors after occurrence, then the system architecture is simple, but the ability to prevent future errors is limited
Solution Approach 1:
The system performs preliminary actions by training the AI model on historical data before errors occur in production environments. The model learns from past errors and their resolutions, enabling it to predict and prevent future errors before they impact users, thereby improving reliability while managing architecture complexity through offline training.
Solution Approach 2:
The system segments the error handling process into distinct phases: data collection, model training, error detection, and resolution recommendation. This segmentation allows the complex prevention capability to be built through modular components, improving reliability without overwhelming system architecture complexity.
4Measurement precision
If accessibility data is collected from multiple devices, then the AI model training data quality improves, but the data collection complexity and processing requirements increase
Solution Approach 1:
The system uses a universal data collection framework that gathers accessibility data, telemetry, and state information from multiple devices using the same methodology. This multi-functional approach improves training data quality by incorporating diverse sources while managing collection complexity through standardized processes applicable across all devices.
Data Source
AI summary
Examples provide an electronic device including at least one electronic processor configured to request first accessibility data associated with a first user interface (“UI”) displayed by a first device and including at least one of (a) information identifying one or more UI elements in the first UI or (b) information identifying one or more UI events in the first UI; train, based on at least the first accessibility data and an error state associated with the first device, an artificial intelligence (“AI”) model; request second accessibility data associated with a second UI displayed by a second device and including (a) information identifying one or more UI elements in the second UI or (b) information identifying one or more UI events in the second UI; and detect, based on at least the second accessibility data and the AI model, the error state on the second device.