The challenge
Società Dolce, a major provider of a wide range of services, uses the ticketing system GLPI to monitor and resolve internal requests. However, the growth in the volume of tickets and their complexity highlighted some problems:
- Overload of IT operatorswhich took too long in the manual handling of tickets.
- Excessive use of resourceswith a negative impact on costs and efficiency.
- Extended response timeswhich slowed down internal processes and decreased user satisfaction.
The main problem lay in the need to improve ticket management through theautomation of responseswithout compromising the quality of the solutions provided. Unlike other companies that merely integrate pre-configured AI services, Synextya has developed a customised solution, tailored to the specific needs of Dolce companies.
The solution
Synextya designed aadvanced AI integration to optimise the management of tickets, through a tailor-made approach that would provide greater efficiency and control for operators.
The solution was implemented in several stages:
- API integration between GLPI and AI services: Creation of a bridge panel to receive tickets from GLPI, test and modify AI-generated answers.
- Choosing the most suitable AI platform: Evaluation of the main AI services available (Amazon Bedrock, Amazon Lex, ChatGPT, Claude 3, etc.) and selection of the most effective one for the needs of Sweet Society.
- Training of the selected modelAI customisation through proprietary data and historical tickets to ensure contextualised and accurate answers.
- System optimisation: Refining prompts and reducing the use of computing resources to improve performance and contain costs.
Project Steps
Step 1: API integration and test panel creation
- Objective: Connect GLPI to AI services via API and provide operators with a panel to test and verify the generated responses.
- Results:
- Creation of a management panel to view, edit and test tickets.
- Implementation of a system to test different AI configurations with real or simulated tickets.
- Generation of draft replies that can be reviewed before sending.
Step 2: Choosing the most suitable AI platform
- Approach: Test the main AI platforms available (Amazon Bedrock, Amazon Lex, Claude 3, GPT, etc.), evaluating the models that best meet the use cases of Dolce Societies.
- Results:
- Identification of the most suitable AI platform for accuracy, efficiency and scalability.
- Implementation of a system allowing operators to select the optimal AI configuration for each ticket.
Step 3: Training the selected model
- Objective: Provide the AI platform with specific information on Società Dolce to improve the quality of responses.
- Results:
- Use of internal documentation and historical tickets to train the AI system.
- Introduction of a human feedback system to continuously refine the answers provided.
Step 4: Fine-tuning and resource optimisation
- Activities performed:
- Reduction of tokens used in prompts, eliminating superfluous terms to lower operational costs.
- Creation of specific prompts based on real examples to improve the consistency and accuracy of responses.
- Results:
- Optimised calculation costs without compromising response quality.
- Increased speed and accuracy in generating AI responses.
The results obtained
Thanks to the integration of AI, Società Dolce achieved numerous benefits:
- Reducing the workload for IT operatorswhich can now concentrate on more strategic activities.
- Improving productivitywith faster and more precise responses to tickets.
- Reducing operating coststhrough the optimisation of AI processes and resources.
The management and test panel will continue to be used for further improvements and tests in the future, ensuring that the system remains up-to-date and performing.
Conclusion
This project demonstrates how advanced and customised AI integration can transform a complex business process into an innovative and efficient solution. Thanks to the collaboration between Synextya and Società Dolce, automation is not only synonymous with innovation, but also with increased efficiency, reduced costs and improved service quality.
