Ask a maintenance engineer at a refinery or a gathering station where the answer to a failing pump lives, and the reply is usually 'somewhere in the system'. Work orders, OEM manuals, inspection reports and lessons learned pile up for decades. They are rarely searchable from the field, and almost never in the language the engineer thinks in.
Why the knowledge stays locked
For a national oil company in the Gulf, critical asset data sat in the maintenance system and in more than 500,000 technical documents, spread across 25+ operational systems. Field engineers needed to ask in Arabic, by voice, in noisy plants — and no existing tool handled multilingual technical terminology.
- Asset history split across many operational systems
- Hundreds of thousands of documents nobody can search from the field
- Questions asked in one language, documentation written in another
- Decisions waiting days for someone to compile the evidence
What we built
We delivered a multilingual LLM platform integrated with the maintenance system and enterprise data. Engineers ask by voice in Arabic or English and get answers in under five seconds, with visualisations generated on demand and translations in both directions, so field and office work from the same facts.
From question to decision
What changed for the asset-management team
| Measure | Before | After |
|---|---|---|
| Time to a maintenance decision | Days | Minutes |
| Answer to a field question | Search and wait | Under 5 seconds, by voice |
| Languages | One at a time | Arabic and English |
| Equipment lifespan | Baseline | Extended by 37% |
Note: Delivered results from the engagement.
Source: DaasLabs, “Delivered for energy, pipelines and heavy industry: energy case study” (2026)
Where to start
Pick one asset class where failures are costly and history is rich — rotating equipment, compressors or transformers — connect the maintenance system and the documents behind it, and measure the time from question to decision. Extend to predictive insights once engineers trust the answers.