Fill in flight delay compensation claim forms
An agent checks what happened to a flight, works out compensation owed under the law, and fills in the airline's claim form, stopping before Submit. Task: The agent takes a flight number and date, converts airline codes (e.g., AI to AIC), and scrapes gate times from FlightAware, cross-checking with AirHelp. It reads live regulations from legislation.gov.uk and validates compensation amounts against a built-in rules table. Finally, it fills the airline's claim form in a cloud browser using Playwright, injecting a script to block form submissions and data transmission.; tools: custom-built or unnamed agent. Human role: Source evidence is unavailable regarding specific human roles or oversight steps.; limitations: Requires a capable model (gpt-5-mini) as smaller models mis-fill fields. Costs ~14 Anakin credits per claim. The agent must inject a lock to prevent submissions, as regex-based button blocking failed on 8 of 10 trap pages. Human review is required before final submission..
Tools
custom-built or unnamed agent
Published on Agent Work Atlas:
Overview
An agent checks what happened to a flight, works out compensation owed under the law, and fills in the airline's claim form, stopping before Submit.
Workflow
The agent takes a flight number and date, converts airline codes (e.g., AI to AIC), and scrapes gate times from FlightAware, cross-checking with AirHelp. It reads live regulations from legislation.gov.uk and validates compensation amounts against a built-in rules table. Finally, it fills the airline's claim form in a cloud browser using Playwright, injecting a script to block form submissions and data transmission.
Reported result
For Air India flight AI162, the agent identified a 207-minute delay and calculated an entitlement of £260 to £520. A full claim process takes two to five minutes and costs approximately 14 Anakin credits. Smaller models like gpt-4.1-nano failed by mis-filling fields, while gpt-5-mini performed correctly.
Human role
Source evidence is unavailable regarding specific human roles or oversight steps.
Limitations
Requires a capable model (gpt-5-mini) as smaller models mis-fill fields. Costs ~14 Anakin credits per claim. The agent must inject a lock to prevent submissions, as regex-based button blocking failed on 8 of 10 trap pages. Human review is required before final submission.
Try it
Act as a flight delay compensation agent. Input: [Flight Number], [Date], [Passenger Details]. Steps: 1. Convert airline code to ICAO format. 2. Fetch gate times from FlightAware and cross-check with AirHelp. 3. Read applicable regulations from legislation.gov.uk and calculate compensation using a rules table. 4. Open the airline's claim form in a cloud browser. 5. Fill the form fields accurately. 6. Inject a script to block HTMLFormElement.submit and fetch requests to prevent accidental submission. 7. Stop before clicking Submit. Output: A filled form screenshot and a calculated compensation range. Verify: Ensure delay is measured at the gate, not takeoff. Check that the model does not invent legal quotes. Human confirmation required before any real submission.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.