AI Mobile App Development in Doha, Qatar

Energy / Oil & Gas

A Qatar-based energy services company (QAR 850M revenue, serving QatarEnergy and international oil companies operating in Qatar, 2,200 field technicians) conducted 45,000+ equipment inspections monthly across onshore and offshore facilities. The problems: (1) Manual inspection: technicians completing paper inspection forms — 127 checkpoints per inspection, handwritten notes in Arabic, photos taken on personal phones and emailed separately. Average inspection time: 2.5 hours (including 45 minutes of paperwork). The paper forms: transcribed into SAP by office staff — adding 30 minutes per inspection and introducing transcription errors (estimated 8% error rate). (2) Inconsistent quality: inspection quality varied significantly between technicians. Experienced technicians identifying subtle equipment degradation (hairline cracks, early-stage corrosion, bearing wear indicators) that junior technicians missed. No standardised method for assessing severity — one technician's "minor corrosion" was another's "significant degradation." (3) Offline requirements: offshore platforms (4 platforms in Qatar's North Field) had limited connectivity — satellite bandwidth shared across the platform. The existing SharePoint-based inspection system: unusable offshore. Technicians reverting to paper when offshore and submitting inspections when they returned onshore — creating 7-14 day reporting delays. (4) Photo management: 350,000+ inspection photos monthly — stored on individual phones, emailed to office staff, and manually linked to inspection records. Photos frequently lost, mislabelled, or unlinked. Finding the photo evidence for a specific inspection from 6 months ago: effectively impossible.

We built an AI-powered field inspection app over 14 weeks: (1) AI-assisted inspection: the app guiding technicians through the 127-checkpoint inspection with camera-based AI assessment at each visual checkpoint. The technician pointing their camera at the equipment — the on-device AI model (Core ML / TensorFlow Lite) identifying: equipment type (valve, pipe, vessel, pump — from 45 equipment categories), surface condition (clean, surface corrosion, pitting corrosion, coating failure, mechanical damage), and severity scoring (1-5 scale — consistent across all technicians because the AI model applies the same criteria regardless of technician experience). The AI assessment: presented as a recommendation that the technician confirms, adjusts, or overrides — the technician retains professional judgment, but the AI provides a consistent baseline. For complex assessments (where the on-device model confidence is below 85%): the app sending the image to a cloud model for enhanced analysis (comparing against a training dataset of 120,000+ labelled inspection images accumulated over 2 years of the client's inspection history). (2) Arabic voice notes: speech-to-text for technician observations — technicians speaking notes in Gulf Arabic (the natural language for Qatari and many non-Qatari technicians working in Qatar), the app transcribing to Arabic text and attaching to the inspection record. For technical terms that mix Arabic and English (extremely common: "valve ده فيه leakage بسيط" — "this valve has a slight leakage"): the speech-to-text model handling code-switching without requiring the technician to stick to one language. (3) Offline-first architecture: the entire app functioning offline — inspection forms, AI models, and photo storage all on-device. When connectivity is available (onshore, or when the offshore satellite link is free): the app syncing completed inspections, photos, and AI assessments to the cloud backend. Sync: intelligent prioritisation — safety-critical findings synced first, routine inspections queued for batch sync. Conflict resolution: for the rare case where an inspection is edited both on-device and on the server — the device version always wins (the technician on-site has the most current information). (4) SAP integration: completed inspections automatically creating maintenance notifications in SAP PM (Plant Maintenance) — eliminating the office transcription step. The AI severity scoring: mapping to SAP priority codes (AI severity 4-5 → SAP Priority 1 "immediate," severity 3 → Priority 2 "urgent," severity 1-2 → Priority 3 "normal"). Photos: automatically linked to the SAP functional location and equipment record. (5) Analytics dashboard: web-based dashboard for inspection managers — showing: inspection completion rates by facility and technician, AI-identified trends (increasing corrosion severity across a specific pipeline section, recurring issues in a particular equipment category), and predictive alerts (the AI model identifying equipment where condition is deteriorating faster than expected — suggesting inspection frequency adjustment).

Inspection time: from 2.5 hours to 1.4 hours per inspection (44% reduction). The time saving: primarily from eliminating paperwork (camera-based AI replacing manual checkbox forms) and voice notes replacing handwritten observations. At 45,000 inspections monthly: 49,500 technician-hours saved per month. Consistency: AI severity scoring providing uniform assessment — inter-technician scoring variability reduced from ±1.8 severity points (manual) to ±0.4 points (AI-assisted). The experienced-vs-junior technician quality gap: narrowed significantly (junior technicians now guided by AI that embodies the assessment patterns of the most experienced inspectors). Three critical equipment conditions in the first 6 months: identified by the AI that the assigned technician initially scored as "minor" — the AI's higher severity assessment triggering review by senior engineers, who confirmed the AI's assessment and ordered immediate intervention. One of these: a pipe wall thinning condition that, if undetected, would have progressed to a leak within 4-6 months — estimated cost avoidance: QAR 2-5M (unplanned shutdown + environmental remediation). Offline reliability: offshore inspections now submitted same-day (vs. 7-14 day delay with the previous system). 100% of offshore inspections captured digitally — zero paper fallback needed. Photo management: 350,000+ monthly photos automatically linked to inspection records, equipment, and functional locations. Finding photo evidence for any inspection: 5 seconds (search by date, equipment, technician, or severity — vs. "effectively impossible" previously). SAP integration: office transcription eliminated — saving 30 minutes per inspection × 45,000 monthly = 22,500 hours/month of office staff time. Transcription error rate: from 8% to 0.3% (the remaining errors: from technician data entry, not transcription). Development cost: QAR 520,000 (14 weeks). Annual value: QAR 4.5M (technician time savings + office staff elimination + critical detection value). The pipe wall thinning detection alone: potentially saving QAR 2-5M in a single incident. ROI: development cost recovered within first 6 weeks from technician and office efficiency gains.