Kela uses artificial intelligence (AI) responsibly and transparently in its customer service, when making decisions on benefits, and for administrative tasks.
What is the AI register?
Information on all AI systems in production use at Kela is stored in the Kela AI Register. The register helps our customers and stakeholders to understand how Kela uses AI.
Kela AI Register
Kela has a chatbot named Kela-Kelpo on its website and in OmaKela. The chatbot provides service in Finnish and Swedish and helps customers find answers to questions they have concerning benefits.
Two chatbots are available on the Kanta Services website: The Kanta chatbot helps citizens in the section titled “citizens” on the website. Depending on their area of expertise, the chatbot for health and social services professionals operates on certain kanta.fi pages for health and social services professionals.
All answers given by the chatbots have been written by a team of specialists at Kela. This means that the AI does not create any of the answers.
AI is used to understand the user's question and to direct it to the correct pre-written answer. Based on sample questions, the AI model learns to identify the topic to which each question should be directed. This means that the chatbots can answer questions even when the customer phrases them differently than in the example questions.
The chatbots have been implemented using the service Boost.ai.
Customers have the right to request transcripts of calls that they have made to Kela’s customer service. We use AI assisted speech technology when we transcribe the calls, i.e. convert the recordings to text.
The transcription tool has been implemented in a secure manner: the language model runs in Kela’s own IT environment, and customer data are not used in the development or training of the tool.
Kela analyses freely written customer feedback with the help of a language model, which is a type of AI. The analysis tool helps Kela identify key themes and areas for development in customer feedback. The results are published as part of Kela’s reporting on the customer experience.
The language model runs in Kela’s own IT environment, and customer data are not transferred outside Kela.
Kela uses machine translation to translate documents submitted by customers and partner organisations into Finnish. The translation tool is primarily used in the processing of international matters and debt recovery cases. Machine translation is used as a support tool by Kela’s specialists, and a human being always checks the translation before it is used in decision-making.
Kela’s translators will also gradually start using the translation tool. It will be integrated into the translation memory software already used by Kela’s translators as a plugin The translation tool will suggest Swedish and English translations, which are then approved, rejected or edited by the translators.
The translation tool uses technology by the system supplier Tilde.
Kela uses AI to pre-process attachments submitted by customers and partner organisations in OmaKela. The system for the pre-processing of attachments automatically prepares attachments so that they can be processed by benefit specialists.
For example, the system
- identifies the type of attachment
- converts the documents to the file format required by the benefit processing system
- edits the attachments for easier processing (converts colour images to black and white and rotates the files so that they are the right way up).
All processing of attachments takes place in Kela’s own IT environment. Automatic pre-processing speeds up benefit processing and ensures that documents are in the correct format for the benefit specialists.
Kela monitors the use of its information systems and network traffic with the help of automated monitoring tools that can identify abnormal usage patterns. The monitoring is based on statistical methods and predefined rules, and partly on AI (machine learning).
The monitoring tools analyse system log data and help ensure that customer data remain secure and that systems are used appropriately.
One of the tools available to those working at Kela is the Microsoft 365 Copilot AI assistant. Copilot helps Kela’s staff in their day-to-day tasks, such as writing different kinds of texts, finding information and condensing content.
Copilot operates in Microsoft’s cloud service in a Microsoft 365 environment managed by Kela. The tool is used in accordance with Kela’s data protection and information security protocols.
Kela has developed an AI assistant called Digiapuri to help its staff in the use of digital tools. Digiapuri answers questions regarding Microsoft applications such as Teams, Outlook and SharePoint.
The assistant was made using Microsoft Copilot Studio. It retrieves the answers to questions posed to it by Kela staff from Kela’s internal instructions and Microsoft’s public guides. If Digiapuri is unable to find the answer or if the problem is complicated, it will direct the user to contact Kela’s IT support.
Digiapuri is an internal tool intended for Kela’s staff, and it operates in a Microsoft Power Platform environment managed by Kela.
Principles for the responsible use of AI at Kela
The principles for responsible use of AI at Kela have been updated to meet the requirements set out in the EU Artificial Intelligence Act as well as Kela’s strategic objectives. These principles govern all use of AI at Kela and ensure that customers and staff can place their trust in services and tools that employ AI.
We abide by the following principles:
- We use AI openly and transparently.
We tell people why, where and how we use AI as part of our work in a way that is clear and easy to understand. We maintain a public list of the AI solutions we employ and their purposes. - We use AI to benefit humans and society.
We aim to promote equity and improve the customer experience of all population groups across Kela’s services. We use a variety of methods to engage different groups of users and specialists in the development and testing of solutions that utilise AI before implementing these solutions. - We ensure that our use of AI is sustainable and inspires trust.
A human is always responsible for the use of AI and its impacts. We build trust by ensuring that our AI solutions operate consistently. We respect people’s right to privacy. - We recognise the impacts of using AI and manage the associated risks.
We assess the impacts of AI applications, programs and systems and manage the associated risks throughout their life cycle. We look critically at the data and algorithms to identify and correct biases. - We work together to develop AI solutions.
We share our expertise as part of broader societal development. We keep up with the times insofar as the development of AI and technology is concerned.
Frequently asked questions about the use of AI at Kela
AI makes customer service faster and more convenient, for example by connecting a customer’s question to the correct answer in our chatbot. However, AI will not replace humans, and the customer can always get help from a human Kela employee.
You can use our feedback form to give us feedback. We use the feedback we get in the development of AI systems and to ensure we operate responsibly.
Kela assesses the impacts and risks of applications, programs and systems that utilise AI. This is done throughout their life cycle. Before implementation, each AI system is subjected to a data protection and information security impact assessment, and if needed, to an ethical assessment. Feedback is also used to develop the solutions. Kela’s goal is to ensure that all solutions it uses prioritise the fair and equal treatment and wellbeing of all people.