evidoria

← Back to browse

Good practice Imported

Seoul Data Hub's Generative-AI Chatbot for Natural-Language Open-Data Discovery

South Korea · Seoul · See the South Korea profile · See the Seoul profile

Evidence: Observational / pre–post Top 47% 60/100 · Ask Evidence Copilot about this practice

Seoul's Data Hub uses a retrieval-augmented generative-AI chatbot to let residents search 8,100+ open datasets in plain Korean; in its first ~2 months it logged 67,300 views and 13,800 active users, with the city openly flagging that answer precision still needs work.

67,300
Cumulative chatbot views during pilot (to 20 Jan 2025)
13,800
Active users during pilot (to 20 Jan 2025)
~80 %
Surveyed user satisfaction
Seoul Data Hub's Generative-AI Chatbot for Natural-Language Open-Data Discovery

Details

Maturity
Scaling
Promoter
Seoul Metropolitan Government, Digital City Bureau, Data Strategy Division
Period
2024–present
Keywords
open data, generative AI, RAG, citizen services, digital government

Context

The Seoul Metropolitan Government's Data Strategy Division launched a generative-AI chatbot inside its Seoul Data Hub open-data portal, letting residents and businesses query the city's roughly 8,100 open datasets in plain Korean instead of navigating manual catalogue menus.

Activities

The chatbot uses retrieval-augmented generation (RAG) -- retrieving matching datasets from the city's catalogue first and grounding its response in that retrieved content, a design the city says was chosen specifically to suppress hallucinated answers -- and began as a pilot in November 2024 before officially opening in February 2025, billed as the first natural-language public-data service among Korean local governments.

Results

By 20 January 2025, during the pilot period, the chatbot had logged 67,300 cumulative views and 13,800 active users, with roughly 80% of surveyed users satisfied and about a third returning five or more times a month.

Conclusions

Seoul's own reporting is candid that the system does not yet return pinpoint-accurate results for every query, describing improving retrieval precision as ongoing work -- a caveat independent Korean IT press coverage of the launch corroborates without disputing the underlying usage figures.

Implementation
Indicative cost
Low (< €50k) — Built on top of an existing municipal open-data portal rather than as new infrastructure, keeping costs closer to a software/integration project than a large capital investment.
Time to results
Short (< 1 year) — Moved from pilot (November 2024) to official public launch (February 2025) in about three months -- a short, fast-iterating rollout.
Staffing & skills
Seoul Metropolitan Government, Digital City Bureau, Data Strategy Division

Conditions for success

  • an existing large open-data catalogue (roughly 8,100 datasets) for the chatbot to retrieve from
  • a retrieval-augmented generation design chosen specifically to ground answers and suppress hallucination
  • a pilot period used to gather usage and satisfaction data before the wider public launch

Common failure modes

  • the city itself acknowledges the chatbot does not yet return pinpoint-accurate results for every query
  • usage and satisfaction figures are self-reported over a roughly two-month pilot window with no independent evaluation yet

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.

Attachments

Similar practices you may find useful