Client’s Request
The client needed to improve the capabilities of its internal help desk, which was already managed by a chatbot.
The main challenge was to automatically direct the problem to the right person or department, so as not to waste the user’s time and respond quickly and accurately to their requests.
Hermes’ Intervention and Involved Talents
Since there was no need to hire external resources and the project had to be developed internally, respecting the specific needs of the company, Hermes made its digital talents available to the client through the WINGS mode.
The talents involved were a Data Scientist and a DevOps. The Data Scientist was responsible for merging three different datasets, translating from various languages to Italian, performing final dataset cleaning and processing, selecting a machine learning model, and training the model. The DevOps, on the other hand, was responsible for creating and deploying an API that allows access to the model and integration with LabelStudio to create a feedback loop.
Offered Solution
After conducting a process analysis and technological assessment, Hermes proposed the development of an algorithm to classify problems based on conversations with a chatbot: the user describes their request to the chatbot, and the description is used by the algorithm to classify the problem so that the chatbot can direct the user to the right person or provide an immediate response. Hermes’ talents also created a feedback loop to optimize the algorithm over time.
The technologies used to develop this algorithm included Python (development language), Sklearn (machine learning library), Fastapi (REST API framework), Label Studio (annotation tool), and Docker (deployment tool).
Results
Thanks to Hermes’ intervention, the following results were achieved:
- With artificial intelligence, the most common problems are solved directly by the chatbot without the need for specific personnel.
- Increased speed in problem resolution.
- Improved efficiency in directing non-standard problems.
- Increase in cases handled in a year (up to 15,000).
+50% increase in standard cases resolved.
