What happened
Nerima Ward in Tokyo collects residents' tax and national health insurance premiums, with arrears of over 2 billion yen a year and about 35,000 people in arrears. Its revenue section proposed an AI to Fujitsu Japan, which supplies the ward's core system, and the two developed the 未納対策支援AI (unpaid-tax support AI). The ward announced on 27 March 2024 that it would start on 1 April 2024, as the first in Japan. The AI learns from data on people in arrears and from veteran staff know-how. It proposes which bank or credit union branches to query for a person's assets, and a second system estimates case difficulty and matches cases to staff skill. Tokyo Shimbun reported on 30 October 2026 from an interview with the section's head. The asset-search work per case fell from about 30 minutes to about 3 minutes, and annual arrears fell by more than 400 million yen. Staff still decide the queries and any seizure. The figures are the ward's own, with no outside audit, and the article is paywalled beyond its opening.
Sources
-
News report, by authors with no tie to the system.
奥野斐 (2026). 練馬区の税滞納が4億円減った データ分析とベテランの技学習...AIで税務が劇的進化. Tokyo Shimbun, 30 October 2026.
滞納額は年4億円超も減った。
導入前には滞納案件1件に平均約30分かかった作業は、約3分まで短縮した。
-
Press release, by the organization that used it. Context or contrary evidence.
練馬区 (Nerima Ward) (2024). 練馬区と富士通Ｊａｐａｎで共同開発した"未納対策を支援するＡＩ"を全国で初めて開始します. Press release, 27 March 2024.
過去の財産調査の作業記録をＡＩに学習させて難易度を推定し
Details
AI system
- Workflow
-
The AI was trained on records of past asset searches and on veteran staff know-how, and it proposed candidate banks and branches to query for each case. Staff judged the proposals, queried the institutions and seized assets where found.
Money
- Kind of value
-
The fall in the ward's annual arrears of residents' tax and national health insurance premiums
- Amount
-
4億円超
- Recipients and payers
-
Nerima Ward
Follow-up
- Conditions for replication
-
The ward must keep feeding new cases to the AI, as its release says it will. Staff must still judge the proposals and carry out the queries and seizures.
Entry details
- Sector
-
Tax compliance.
- Purpose
-
Nerima Ward and Fujitsu Japan built the unpaid-tax support AI to propose the best places to search for a delinquent taxpayer's assets and to match cases to staff by difficulty.
38 more questions not yet checked
model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, currency and prices, period and unit, realization, gross or net, included costs, comparison, method, return and payback, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this benefit
Peter H. Vartanian. (2026). Nerima's unpaid-tax support AI cuts asset searches to 3 minutes (AIBD-1191). Artificial Intelligence Benefits Database, edition of 2 November 2026. https://petervartanian.xyz/aibd/records/AIBD-1191.html (archived at https://doi.org/10.7910/DVN/AADSII, version 9.0)