Stefan Hüsemann
Partner
FDFA: Faster customer support through helpdesk mailbot
AI innovation project with concrete benefits for the Swiss Federal Department of Foreign Affairs (FDFA)
November, 2020
The Swiss Federal Department of Foreign Affairs (FDFA) has improved its IT support process with the help of Artificial Intelligence (AI) as part of an innovation project. The most important goals were achieved: Customer support requests are resolved more quickly and the IT FDFA has built up AI know-how in a specific use case.
Challenges
The FDFA's IT department, with around 100 employees, provides IT services around the clock for the FDFA around the world (e.g. for Swiss embassies, consulates and the Swiss Agency for Development and Cooperation). In doing so, IT is confronted with the following challenges, among others:
-
Delivering a wide range of IT services efficiently to customer satisfaction.
-
Besides the daily business, keep your eyes open to use innovative technologies in suitable applications.
-
Continuous digitization of e-government processes.
In concrete terms, this means that the IT helpdesk, for example, with a maximum of 14 employees, has to process around 5000 queries per month, some of them complex, in 24/7 operation. The support processes involve many manual steps and transfers between various people and systems.
Impacts
As a small IT service provider in the federal administration, the FDFA IT must differentiate itself through special customer proximity, efficiency and innovation. If this is not achieved, the FDFA cannot perform its tasks, which ultimately has an impact on the satisfaction of globally distributed FDFA stakeholders.
Solution
In order to tackle challenges in the FDFA IT and its helpdesk (HD), an AI innovation project was launched. The resulting HD Mailbot automates various tasks in the IT support process.
-
Analysis, allocation & prioritization
Incoming support request mails are analyzed by the bot using AI (Natural Language Processing - NLP), assigned to an affected IT service, prioritized and assigned to a support group.
-
Support ticket
The mailbot then automatically creates a ticket in the support system.
-
Automated assistance
In addition, the mailbot identifies existing assistance with relevant instructions based on similar cases. The bot sends these back to the customer as an initial solution suggestion. Ideally, the customer can then solve his problem independently.
The solution was largely created by FDFA IT. ipt was supported by AI experts throughout the entire project. Thus, an optimal know-how transfer could be achieved and the successful implementation was ensured.
Business Value
The following business value was generated by the AI innovation project and the resulting helpdesk mailbot.
Click here for an explanation of the ipt Value Spider – Make technology valuable
The customers of the FDFA helpdesk have the benefit that their requests are answered faster (Customer Value Case). The foundation for process automation in the helpdesk environment has been laid, leading to an increase in efficiency in the medium term (Infrastructure Case & Efficiency Case). FDFA IT has gained experience with machine learning and built up strategically important AI know-how (Strategy Case).
In figures - this was achieved
of German language mails are analyzed by the mailbot.
service types identifiable by AI algorithm in support requests.
support tickets created automatically per day including service allocation.
e-mails used as training data for machine learning
Excursus: Chatbot prototype for Swiss representation in France
As an introduction to the topic of Artificial Intelligence, FDFA IT created a chatbot prototype for the search of consular information. ipt acted as an enabler in this project and provided the project staff with the know-how for a start in the AI world. In the pilot phase, this bot answers questions from users on the public website of Swiss representations in France. The chatbot is used extensively - in the first 2 months the chatbot has had over 1400 conversations. The continuously collected key figures serve as a basis for assessing the benefits of this solution and for future planning in this area:
- 1400+ conversations in the first 2 months of the pilot phase Approx
- 80% hit rate Approx
- 90% flow-based input and 10% manual text input
- 38 intents with 15 to 70 uterances each
Tech-Box
The following is required from a technological point of view
- Assignment of support requests to IT services and assistance based on historical requests using machine learning algorithms.
- Integration with MS Exchange Mailserver and BMC Remedy IT Service Management System.
- Guarantee of data protection for mail contents.
The following was used
- Data Understanding: Python using Jupyter notebooks.
- Data Preparation: Natural Language Processing Libraries NLTK and Spacy.
- Data Modeling: Machine Learning models from Scikit-Learn and the Deep-Learning Model BERT on Azure Machine Learning.
- A microservice composition of .NET Core and Python Flask components.
- Agile process methodology in the DevOps team.
We were on the project
Please contact us if you have any questions.
