AI assistant piloted in the city intranet
The autumn experiment of the Espoo Digiagenda involves testing an AI assistant in the city’s intranet Essi. One aim is to improve the discoverability of information. Another aim is to examine whether artificial intelligence will make work more efficient and to prepare the organisation for using generative AI solutions in the future.
More than 210,000 searches are made in Essi each year to find information and instructions to support work. However, many users find that locating information is slow and difficult. Successful search requires knowledge of Essi’s structure and the use of correct keywords, and the current search does not guide the user.
Three different solutions for testing
The experiment will be carried out in a test environment. The test will cover the current intranet search, a tailored GPT AI assistant and MS Copilot.
The group testing the tools includes a total of some 30 employees from the Digital Service Development and Knowledge Management department, and communications, finances, HR and the project office. The participants are immediate supervisors, developers and experts.
“We have selected different use cases for the experiment. They’ll help us gain insight of the roles, tasks and parts of processes that an AI assistant is most useful for. At the same time, we’ll get comparative data on the strengths and weaknesses of the GPT and Copilot”, says Development Manager and pilot leader Valia Wistuba.
The test environment, whose planning and setup also involves the city’s information management, will be built in September, and the testing will happen between October and December. Then, the results will be analysed and the lessons learned from the experiment will be shared around December and January.
Will AI improve the efficiency of work and the discoverability of information?
“The aim of the experiment is to learn about utilising generative AI, bring awareness to the significance of high-quality source data and to examine its impacts on improving the efficiency of work. At the same time, we’ll compare the benefits and costs of different solutions”, says IT Development Manager Marko Piippo.
At their best, AI assistants save working time and make it easier to prepare materials such as orientation or communication plans by facilitating information retrieval and administrative tasks. Also, when instructions are readily available, the need supervisory work and orientation is lessened. Looking at source data helps to find ways for improving the quality of different kinds of content.
“Experiments also always offer an opportunity to develop personnel competence. This experiment will help us assess what kind of competence we’ll need in the future if we want to utilise generative AI for something like administrative tasks on the intranet”, says Piia Wollsten, Director of Digital Service Development.