"Gangnam with energy! Residents with joy! The great transformation of Gangnam!!" — that was the campaign slogan Kim Hyun-gi carried into office when he was inaugurated as Gangnam district mayor under the ninth round of local elections.
That slogan is now taking concrete shape. The district has streamlined reconstruction project approvals and, in the latest example of its push for administrative innovation, a staff member has built a system that gets overpaid taxes back to residents faster.
A district office employee used generative AI to develop an in-house name-matching verification system, which was applied in June to the bulk processing of personal local income tax refunds on comprehensive income. As a result, average processing time fell from five days to two.
Personal local income tax on comprehensive income is a local tax filed and paid alongside the national comprehensive income tax, which aggregates an individual's annual earnings from business, employment, interest, dividends and other sources. Freelancers and instructors, for example, have a portion of their income tax and personal local income tax withheld at source when they receive payment. When final tax liability is calculated the following May — factoring in actual income, deductible expenses and various credits — those who paid more than they owed receive a refund of the difference from their local government.
Each year at the end of June, roughly 70,000 personal local income tax refund records generated from May's comprehensive income tax filings arrive at the district office in one batch. To disburse refunds accurately, staff must confirm that the name of each refund recipient matches the name on the designated bank account. Previously, officials had to manually search through the entire dataset to identify mismatches, a time-consuming process.
To reduce that repetitive workload, a district employee developed the name-matching verification system in-house. The program automatically cross-references the refund recipient's name against the account holder's name and extracts only the records where the two do not match. Staff then need to review only the mismatched entries rather than the full dataset.
The district applied the program to the roughly 70,000 refund records in June. As a result, the average time from receiving the refund data to verifying names, updating the system and submitting payment requests to banks fell from five days to two. Building on that faster internal processing, the district completed all refund disbursements in early July, allowing taxpayers to receive their money sooner than before.
"Refunds represent taxes that residents overpaid, so we have an obligation to return them quickly and accurately," district mayor Kim said. "We will continue to use generative AI and digital technology to improve repetitive administrative tasks and expand the kind of service innovation that residents can actually feel."
seouldream01@heraldcorp.com
