Thailand's NACC deploys an OCR-NLP-ML pipeline to pre-screen incoming corruption complaints, cutting average processing time by 78.6%. A 2025 peer-reviewed evaluation found an OCR F1-score of 81.8%; human reviewers retain full final decision authority.
81.8 %
OCR module F1-score (2025 study)
84.2 %
OCR precision (2025 study)
79.6 %
OCR recall (2025 study)
57.5 %
Document classification accuracy (2025 study)
78.6 %
Average complaint processing time reduction (2025 study, vs manual baseline)
Details
Maturity
Established
Promoter
National Anti-Corruption Commission (NACC)
Period
2024–present
Keywords
anti-corruption, complaint management, document processing, public integrity
Context
Thailand's National Anti-Corruption Commission (NACC) has deployed an AI-powered complaint-screening pipeline combining Optical Character Recognition (OCR), Natural Language Processing (NLP), and supervised machine learning to automate initial document segmentation, allegation classification, and workflow routing for incoming corruption complaints — tasks previously handled manually. Human reviewers retain full authority over final screening decisions; the AI handles the time-consuming pre-processing stage.
Results
A 2025 peer-reviewed study published in PubMed Central documented the system's performance in operational conditions: OCR module F1-score of 81.8% (precision 84.2%, recall 79.6%), document classification accuracy of 57.5%, and an average complaint processing time reduction of 78.6% versus the manual baseline. The 57.5% classification accuracy is reported transparently by the authors as a current limitation; the system is explicitly positioned as a triage aid, not a final decision-maker. Thailand's adoption of AI for complaint handling is noted in Thailand's chapter of the OECD Anti-Corruption and Integrity Outlook 2026.
Implementation
Indicative cost
Medium (€50k–€500k) — OCR-NLP-ML pipeline built for a single agency's complaint intake; no published budget figures.
Time to results
Medium (1–3 years) — Operational since 2024; 2025 peer-reviewed evaluation published documenting performance under live conditions.
Staffing & skills
Thailand's National Anti-Corruption Commission (NACC) operates the pipeline; human reviewers retain full authority over final screening decisions
Conditions for success
Human-in-the-loop design preserves final decision authority for staff while automating pre-processing
Transparent academic disclosure of the 57.5% classification-accuracy limitation supports credible, incremental adoption
Common failure modes
57.5% document classification accuracy is an acknowledged current limitation
No external algorithmic audit or redress mechanism is documented
No replication of the architecture in another jurisdiction is documented
Where it fits
Governance type
national anti-corruption agency
Scale
single national agency
Income level
upper-middle-income
Commonly funded by
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.
Data sources
Where this practice's information was retrieved from, and when.
Ukraine's National Agency on Corruption Prevention automatically risk-scores officials' asset declarations, clearing roughly a third of the 1.05 million declarations …
Armenia's Corruption Prevention Commission uses AI to scrutinize its World Bank-funded Electronic Platform for Declarations, launched Feb 2023, with 31,579 …
Mongolia's Independent Authority Against Corruption uses the AI-powered tuss.io compliance platform to screen state orders, decisions, rules and regulations for …
Andalusia's regional government launched JuntaGPT, an internal generative-AI assistant giving up to 240,000 public employees access to a Gemma 3-based …
