Italy·Public sector·2025 – ongoingENAV
STORY, AI pipeline for air traffic safety investigations
ENAV's AI pipeline for safety investigations cut effort by 60% in testing, the company says
Problem
Safety investigations needed investigators to listen to pilot and controller radio recordings and transcribe them word by word. Coded aviation language had to be interpreted by hand. Transcribing a 10-minute recording took about three times its length.
Approach
ENAV's Innovation Lab built a pipeline with a custom speech recognition engine, because commercial tools were not accurate enough for aviation radio. It separates pilot and controller speech, aligns airline and airport codes with international standards, and uses a large language model with retrieval-augmented generation to draft a report from internal documents and procedures. A controller-investigator checks, corrects and signs each report. The tool was in testing at the Rome, Milan, Padua and Brindisi control centres and at Malpensa, Fiumicino and Venice airports, with a roll-out to all safety offices planned.
Results
Takeaway
European rules limit the work to offline transcription after an event is reported, and a named investigator stays responsible for each report.
Categories
- Size
- Large enterprise
- Industry
- Public sector
- Business function
- Production and qualityInternal work and knowledge management
- Innovation type
- Process innovation
- AI technology
- Generative AI
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
- AI per la gestione del traffico aereo: l'esperienza di ENAVEconomyUp·29 Sep 2025News article
Source summaries
Reports ENAV's Innovation Lab head describing the STORY pipeline at a 23 September 2025 talk. It covers the components, the human sign-off, the test sites and the 60% effort reduction ENAV states. The project start year is taken from the presentation date.
Full details are in the original sources.
