Versia – Automatic Incident Resolution
Versia
Sector: Logistics
Business Case
This NLP application records incidents reported via email and, following classification using various NLP techniques, automatically initiates the appropriate corrective actions. (Project in progress. Scheduled for production launch: 30/6/2023).
Objectives
Reduce human involvement in incident handling. (Operations Management). Reduce procurement problems (stock-outs) due to the lack of efficiency in incident resolution.
Use case
Use LLMs for entity classification and extraction (NER)._x000B_Zero-shot learning._x000B_Create a Bot with Rasa and trained models for corrective action automation.
Infrastructure
Cloud.
Technology
Machine learning and deep learning NLP
Data
Service tickets Private datasets
Resources
Team assigned to the project. Multidisciplinary team: IT, mathematicians/physicists, organisation engineers. GCP, HEROKU, VAULT, RASA, Transformers (HUGGING FACE).
Difficulties and learning
Lack of data._x000B_Absence of resolution protocols._x000B_Absence of written communication protocols (colloquial and slang communication, lack of context)._x000B_Previous absence of an issue tracking system.
KPIs (business impact and metrics of the model)
Number of tickets resolved automatically. Reduction in the number of stock-outs in stores.
Funding
Private. Successful.
Collaborators, Partners
Hitz.
