# Send a photo of a city problem and have it routed to the office that fixes it

People in Thailand report a pothole or a broken streetlight to Traffy Fondue on LINE, with a photo and location. AI sorts each report and sends it to the office responsible. Bangkok has resolved more than 1 million.

Organisation: NECTEC, part of Thailand's National Science and Technology Development Agency (NSTDA), with the Bangkok Metropolitan Administration
Status: Deployed; used by the Bangkok Metropolitan Administration since 2022 and, by July 2026, by 34,786 agencies in all 77 provinces of Thailand; AI summaries, priority ranking and suggested agencies added in July 2024
Place: TH
Categories: Cities & daily life, Government & public services
Try it with: Claude, ChatGPT / OpenAI models, Gemini, Grok
Time to try: 30 minutes

## Why it matters

Traffy Fondue was built by NECTEC, part of Thailand's National Science and Technology Development Agency (NSTDA), and the Bangkok Metropolitan Administration adopted it officially in 2022. A resident sends a photo and a location, and the platform's AI works out what kind of problem it is and which office should fix it, so the resident does not need to know who is responsible. In figures released in March 2026, the city's spokesperson said Bangkok had received 1,231,758 complaints through it and resolved 1,002,604 (81%). In May 2025 the prime minister at the time, Paetongtarn Shinawatra, told every ministry to expand it into the government's main complaints channel. NSTDA said in July 2026 that 34,786 agencies in all 77 provinces use it, up from 1,900 in 2023, and that the average time to resolve a case had fallen from three months to half a month.

## How it works

Residents add the @TraffyFondue account on LINE, or use the mobile app, then share a location, type what is wrong and send a photo; a July 2024 update cut reporting from five steps to three. An AI engine classifies each report from its text and images and routes it to the agency responsible for that area. NSTDA says the AI also summarises each report, ranks its priority, suggests a fix and the office that should handle it, and flags rude wording with a polite alternative. A report that fits no known category is handled by a person, who files it in the right category or raises a ticket to create a new one, according to a study by Thammasat University researchers. Officials post updates on the case, and the resident follows each step on LINE and rates the result in stars. Agencies are scored on response speed, quality of the fix and citizen satisfaction, which NSTDA says is meant to prevent superficial ticket closures. In July 2026 NSTDA listed AI warnings when a resident types too much personal information, uses unsuitable language or attaches an unsuitable image, and a feature that groups reports of the same problem.

## Who it's for

City and district governments, utilities, public hospitals and universities that take fault reports from the public, and civic technology teams building complaint and service-request tools.

## The prompt

```
You are helping me design a service that lets residents report a local problem (a pothole, a broken streetlight, a blocked drain) with a photo and location, and uses AI to sort each report and send it to the office responsible. Do not ask me for any real resident's name, phone number or report.

Our area: [for example, a city of 2 million with 50 district offices and separate water and power utilities]
Channels: [for example, a messaging app our residents already use, plus a phone line for people without smartphones]
Offices and what they handle: [list each office or contractor, the problems it fixes and the areas it covers]

1. Propose 15 to 25 problem categories, each with a plain description, two example reports and the office that handles it in each area.
2. Write the rules the AI must follow when it sorts a report: use the photo, the text and the location; send anything it is unsure about to a named person; never close a report itself.
3. Write the three to five messages a resident receives, from confirmation to the result, in plain language, including how to rate the fix and how to reopen the report.
4. Set privacy rules: what to blur in photos, which personal details to warn about before a report is sent, and who can see what.
5. Design a scorecard for offices that weighs response speed, quality of the fix and resident ratings, so that closing a report without fixing the problem scores poorly.
6. Plan a three-month pilot in two districts: what to measure each week, how to check a sample of closed reports on site, and how to reach residents without smartphones.

Leave placeholders rather than invent offices, rules or targets, and mark everything a responsible official must sign off before launch.
```

## Caveats

The figures come from NSTDA, which built the platform, and from the Bangkok Metropolitan Administration, which uses it; no independent audit of them was found. NSTDA's English release of August 2026 speaks of more than 2 million resolved tickets, while its Thai release of July 2026 counts more than 2 million cases in the system; this card uses neither. In Bangkok's March 2026 figures, 19% of residents' ratings of a fix were unsatisfied. None of the sources checked names the AI models behind the platform or says how often its sorting is wrong. The Thammasat University study found the service is smartphone-based, which shuts out people without a smartphone, stable internet or digital skills, and found no clear digital inclusion policies; its data covered nine months in Bangkok only. Use varies by area: NSTDA says 33 of the 77 provinces run it at full scale. A 2022 agreement let local administrative offices use the platform at no charge until 30 September 2025; no current terms were found.

## Sources

The six NSTDA pages, one of them distributed through ThaiPR.net, are the developer's own, and the March 2026 release quotes the Bangkok Metropolitan Administration's spokesperson. The peer-reviewed study is by Thammasat University researchers who declare no funding or conflicts of interest, but it draws on the platform's data and thanks officials at the Bangkok Metropolitan Administration and the Digital Economy Promotion Agency for their support. No independent evaluation of the AI's accuracy was found.

1. NSTDA (National Science and Technology Development Agency, Thailand): "utilizes artificial intelligence (AI) to automatically classify the issue and route it directly". Thailand's AI-Powered Citizen Report Platform Sets New Benchmark in Digital Governance (https://www.nstda.or.th/en/news/news-years-2026/thailand%E2%80%99s-ai-powered-citizen-report-platform-sets-new-benchmark-in-digital-governance.html)
2. NSTDA (National Science and Technology Development Agency, Thailand): "เวลาการแก้ไขเรื่องลดลงจาก 3 เดือน เหลือ 0.5 เดือน". Traffy Fondue จากแพลตฟอร์มรับเรื่อง สู่ระบบบริหารงานรัฐที่วัดผลได้ (https://www.nstda.or.th/home/news_post/traffy-fondue-government-management-system-20260703/)
3. NSTDA (National Science and Technology Development Agency, Thailand): "โดยดำเนินการแก้ไขแล้วเสร็จ 1,002,604 เรื่อง คิดเป็นร้อยละ 81". ความสำเร็จแพลตฟอร์ม “Traffy Fondue” หนุน กทม. บริหารจัดการเมืองดิจิทัล โปร่งใส ตรวจสอบได้ (https://www.nstda.or.th/home/news_post/traffy-fondue-supports-bangkok-digital-city-20260307/)
4. NSTDA (National Science and Technology Development Agency, Thailand): "การขยายผลการใช้ Traffy Fondue ทั่วประเทศ เพื่อใช้เป็นช่องทางร้องเรียนหลักของรัฐบาล". นายกฯ เน้นย้ำขยายผล “Traffy Fondue” ทั่วประเทศ เพื่อเป็นช่องทางร้องเรียนหลักของรัฐบาล รับฟัง-แก้ปัญหาประชาชนให้ตรงเป้า (https://www.nstda.or.th/home/news_post/expand-traffy-fondue-nationwide-250507/)
5. NSTDA, via ThaiPR.net: "สรุปสาระของปัญหาที่ได้รับแจ้ง คัดแยกประเภทปัญหาจากข้อความ และรูปภาพ". Traffy Fondue เปิดฟีเจอร์ใหม่ ใช้ AI สุดล้ำ บริหารจัดการ ‘เรื่องแจ้งปัญหา’ (https://www.thaipr.net/?p=3497519)
6. NSTDA (National Science and Technology Development Agency, Thailand): "Powered by artificial intelligence (AI) and linked with LINE messaging application". Traffy Fondue platform to enhance efficiency of public services (https://www.nstda.or.th/en/news/news-years-2022/traffy-fondue-platform-to-enhance-efficiency-of-public-services.html)
7. Frontiers in Sustainable Cities (Hansen and Dahiya, Thammasat University): "The backend architecture uses an AI engine to process the complaints"; "did not find any clear digital inclusion policies". Traffy Fondue: a smart city citizen engagement (https://www.frontiersin.org/journals/sustainable-cities/articles/10.3389/frsc.2025.1491621/full)

Fact-checked: 6 Nov 2026

Page: https://escapevelocitysociety.com/picks/thailand-traffy-fondue-ai-city-problem-reports

See Selected Projects: https://escapevelocitysociety.com/picks.md
