Skip to content
An official website of the OECD. Find out more
Created by the Public Governance Directorate

This website was created by the OECD Observatory of Public Sector Innovation (OPSI), part of the OECD Public Governance Directorate (GOV).

How to validate authenticity

Validation that this is an official OECD website can be found on the Innovative Government page of the corporate OECD website.

Artificial Intelligence (AI) for Classification of Requests at the Citizen Services Desk.


The initiative is about the automation in the classification of service desk requests using AI techniques such as Natural Language Processing and open source tools to reduce the costs, optimize resources and improve response times in benefit of the citizens and the public institutions as service providers. The solution is developed by AGESIC (E-Government and Information and Knowledge Society Agency); this is an algorithm based solution in charge of classifying topics on different categories of services, and to link the requests typologies with the users information.

Innovation Summary

Innovation Overview

The solution has successfully automated the classification of 82% of AGESIC's Service Desk requests as it was developed as end-to-end platform. It was in production for over a year and nowadays is running and processing more than 33 thousand requests yearly.

Thanks to a centralized system, the service desk manager reviews the platform's operations in real time which, at the same time, builds and runs different pre-set performance indicators. The platform is supervised by the Desk technicians whom collaborate in the detection of wrongful allocation of cases/requests. The platform is supported by the National Supercomputing Center of Uruguay (ClusterUY) allowing the solution to operate 24/7 using the resources from the Cloud of the Presidency. It is easily replicable to other Organizations that present the same problem, thanks to its open design and use of open source tools.

Innovation Description

What Makes Your Project Innovative?

Before implementing this solution another one was tested using Robotic Process Automation techniques, but was not satisfactory as it required permanent maintenance due to its high variability in the output results. From its encouraging results a new solution was derived using NLP techniques to allow our institution to adapt much better to re-training and reproduction of several parts of its algorithm. The use of performance indicators allows to our technicians to monitor live the results and suggest quick improvements and extensions accordingly to the new needs and evidence generated by the platform.

Innovation Development

Collaborations & Partnerships

AGESIC technicians were fully involved into the design and daily operations of the platform, their job was critical to define the scope of the tool and to build the required indicators to monitor its performance and oversee the platform's evolution.

Users, Stakeholders & Beneficiaries

The users of the system saw improvements in the service delivery due to shorter response times and more accurate responses to their requests/problems as the public officers will have more time to "spare" providing solutions rather than allocating documents. According to out polls, all the stakeholders are highly satisfied with the results achieved so far.

Innovation Reflections

Results, Outcomes & Impacts

The impact has been positive and the good reviews and feedback can prove it. Furthermore, the Agency's management board has been very satisfied with the solution and the participation of all the sectors involved. The innovation is considered as a success story in the use of AI in government and because of its interoperability is easy to replicate and implement.

The correct definition of the indicators and the strong monitoring mechanisms have contributed to give transparency and visibility to the solution, which generated several good practices. The system is constantly evolving based on the monitoring of indicators and the results are systematically reviews.

Challenges and Failures

A key challenge was to demonstrate that the implementation benefits could be achieved without resorting to layoffs and employee absences. Moreover, an initial imperative was not to hinder the operation of the Service Desk system so it had to run in parallel. Another important aspect was to maintain fluid communication between all parties, which made it possible to report on progress, setbacks, adjustments and improvements as they occurred.

Conditions for Success

It was necessary to engage all the technical and political parties in the mission of the organization. Given the possibility of failure and the perceptions on the use of technology, we needed to build a speech based in change management and trust in the process. It was also not easy to convince that Artificial Intelligence was more than a label and had several examples of success stories, in particular when allocating public budget.


The solution has great potential for implementation and replicability. The use of reusable, known and proven open source tools offers guarantees significant savings in times and costs. We have also extensively documented the solution as well as the different stages it went through, which allows us to reduce concerns about the use of technology and the data protection mechanisms of the users. Currently, several Uruguayan government agencies have shown interest in implementing it.

Lessons Learned

- New solutions to old problems are now possible thanks to AI.
- A committed and resilient team and leadership is needed to develop the project.
- Experts, preferably with direct tasks per problem, should be integrated from the beginning to ensure an incremental and controlled development, respecting multiple points of view and needs.
- Adequate planning and proper follow-up are necessary to maintain a good pace of work, respect the deadlines imposed and respect the agreed budget."

Year: 2021


Level of Government: National/Federal government


  • Implementation - making the innovation happen
  • Diffusing Lessons - using what was learnt to inform other projects and understanding how the innovation can be applied in other ways

Innovation provided by:


Date Published:

21 November 2022

Join our community:

It only takes a few minutes to complete the form and share your project.