An Ichthys-specific generative-AI tool converts daily reports into a searchable database and answers natural-language questions with past operating performance and troubleshooting cases. A connected function structures maintenance history by categorizing work descriptions, failure details and affected components.
Company
INPEX
Department
Maintenance & reliability
Record type
company_adoption
Disclosed stage
Developed and deployed at Ichthys; specific go-live date and adoption scale are not disclosed.
Source date
2026-07-07
Source-date precision
day
Source provenance
A · Operator filing
Review disposition
Source-attributed; not an independent validation
Reported result
INPEX reports that searches formerly taking several hours now complete in seconds to minutes; the report graphic gives “seconds to minutes.” It says AI structured approximately 110,000 maintenance records and achieved over 90% classification accuracy. No accuracy metric definition, validation sample, baseline method, error rate or time-study method is provided.
What this does not establish
Operator-reported implementation and process metrics. The reported >90% concerns classification accuracy, not overall answer accuracy. Faster information retrieval and record structuring are not quantified as downtime or production improvements. Engineer approval of search quality does not establish mandatory human review of every response or maintenance action.
A concrete, site-specific O&M use case at INPEX’s operating Ichthys LNG project, with disclosed data scale and workflow measures; separate from the CCS reservoir-simulation surrogate candidate.
Latest Report section states Integrated Report 2025 published July 7, 2026; Back Numbers section dates Integrated Report 2024 July 4, 2025.
Disclosure history
Ichthys AI tool searches operational reports and structures maintenance history
An Ichthys-specific generative-AI tool converts daily reports into a searchable database and answers natural-language questions with past operating performance and troubleshooting cases. A connected function structures maintenance history by categorizing work descriptions, failure details and affected components.
A source grade describes provenance, not independent verification. This is a research profile, not a measure of net return. Report a correction.