PLLuM in mObywatel — Poland's Sovereign-Language-Model AI Assistant for Public Services
Poland
A Polish academic-government consortium built PLLuM, a homegrown LLM, piloted in Częstochowa from August 2025 and rolled into the national …
Kazakhstan · Astana · See the Kazakhstan profile · See the Astana profile
Evidence: Descriptive / self-reported Top 67% 53/100 · Ask Evidence Copilot about this practice
Kazakhstan's eOtinish platform processed over 5 million citizen and business appeals in 2025 (up ~1 million on 2024); NITEC has layered AI assistants and analytics on top to speed responses, though the AI layer's own impact isn't yet independently evaluated.
Following Kazakhstan's 2021 'Hearing State' reform, which directed government agencies to replace a fragmented, agency-by-agency process for handling citizen and business appeals with a single national channel, NITEC JSC (the state IT enterprise under the Ministry of Digital Development, Innovation and Aerospace Industry) built and launched the eOtinish platform on 1 July 2022, giving citizens five ways to submit and track appeals: the eOtinish.kz website, the eGov.kz portal, a mobile app, in-person Citizens' Service Centres, and Nätije reception centres.
NITEC describes the platform as adding 'AI assistants' to speed up responses and 'advanced data analytics' to process unstructured complaint text. A regional commentary describes this analytics layer as already screening millions of complaints to surface acute regional issues and emerging sentiment trends, while Kazakhstan's own official channels describe the AI component more tentatively — as a capability still being built out, with 'AI-based analytical tools' and an updated Unified Knowledge Base named as near-term priorities rather than completed, evaluated features.
By the end of 2025, eOtinish had logged more than 5 million appeals for the year alone — about 4 million from individuals and 1 million from legal entities — roughly 1 million more than in 2024. Every appeal can be tracked online; citizens can rate the response, escalate as a formal complaint, or file a claim in court.
No independent evaluation of the AI layer's accuracy, effect on response time, or impact on resolution quality has been published — its measured impact should be treated as unproven even though the underlying platform's operational scale is well documented.
National / regional programmes
Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.
Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.
Where this practice's information was retrieved from, and when.
Poland
A Polish academic-government consortium built PLLuM, a homegrown LLM, piloted in Częstochowa from August 2025 and rolled into the national …
Mauritius
Mauritius' Central Informatics Bureau runs DIVA, an AI chatbot answering citizen queries for eight ministries via web, WhatsApp and app …
Honduras
Honduras's Policía Nacional and Secretaría de Seguridad, with UNDP's Infosegura project, launched MARIA in April 2024 — a free 24/7 …
Egypt
Egypt's CAOA and USAID launched KMT in February 2024, the civil service's first AI chatbot, built on a 50,000-question knowledge …
Open full copilot Grounded in cited practices — always check the sources.