OTT2 is an AI-powered decision support system that helps Estonia’s Unemployment Insurance Fund prevent long-term unemployment. It combines data from public registries to assess each jobseeker’s employment prospects in real time, enabling counsellors to identify those at highest risk and focus support where it is most needed. With above 98% forecast accuracy, it eliminates laborious and error-prone manual work and it enables a policy pilot to dedicate counselling resource to high-risk customers.
Initiative overview
Estonia’s public employment service faces the same challenge seen worldwide: allocating limited support to jobseekers with very different needs. Before 2020, counsellors relied on rules of thumb, personal experience and the manual profiling of roughly 45 indicators ranging from demographic characteristics and work history to health status and local labour‑market conditions. In practice, this volume of information was difficult to process consistently, making it hard to reliably identify people at risk of long‑term unemployment and in need of more intensive support.
This changed with the introduction of OTT1 in 2020, Estonia’s first data‑driven decision‑support tool for counsellors. Using a machine‑learning model, it assessed each jobseeker’s employment prospects at the point of registration. Over several years of operational use, the Estonian Unemployment Insurance Fund (EUIF) strengthened the model, built organisational confidence and gathered the evidence needed to expand its approach. OTT2 represents this next step. Whereas OTT1 produced a single assessment at registration, OTT2 generates dynamic, regularly updated assessments that reflect changes in a jobseeker’s situation—such as completing training or missing an assigned activity—so counsellors can make decisions using the most current information. The model is retrained periodically on new data but does not change autonomously, ensuring stability, transparency and auditability in a high‑stakes service. OTT2 also explains the main factors shaping each prediction and how these evolve over time, helping counsellors interpret results and offer more personalised support.
OTT2 builds on the strong performance of its predecessor, which achieved over 95% prediction accuracy. By updating assessments as circumstances change, accuracy is expected to rise above 98%; this is being tested in a 2026 pilot. The pilot also introduces a targeted service change: the approximately 20% of newly registered jobseekers identified as lowest‑risk will no longer receive mandatory counselling, allowing counsellors to focus more time on individuals at higher risk of long‑term unemployment. Even small reductions in unemployment duration have significant fiscal impact—one day less is estimated to save around 3.8% of annual expenditure.
Other relevant details
Introducing AI into core public‑employment workflows brought technical, organisational and cultural challenges. Early experience showed that tools developed without counsellor involvement led to trust issues and uneven adoption, with newer staff more likely to use OTT1 than more experienced colleagues. To address this, EUIF expanded user engagement through seminars, webinars and clear guidance materials. For OTT2, counsellors were involved from the outset through interviews and feedback loops, ensuring the design reflected real operational needs. The main lesson is clear: end‑user participation from the start is essential. Building trust and usability early is far more effective than attempting to fix resistance after deployment. EUIF now treats early user involvement as a standard requirement for any AI introduced into public services.
OTT has been developed in close collaboration with academic and private sector partners. The University of Tartu contributed research expertise throughout the process. Nortal has been involved from the outset, supporting the identification of use cases, providing data science and technical capabilities, and contributing to the service design underpinning the April 2026 pilot. The initiative has also received national recognition. In 2021, OTT was awarded Best Data-Based Digital Service in the Estonian public sector. The tool is used in daily work by counsellors, supporting more consistent decisions on how to prioritise support and allocate resources.
From April 2026, EUIF will pilot removing mandatory counselling for jobseekers with the strongest employment prospects. This reflects the high accuracy of OTT2 and growing confidence in using AI to guide service allocation. The goal is simple: allow low‑risk jobseekers to progress independently and redirect counsellor capacity to those who need support the most, improving both service quality and system efficiency.
