SOCAR scales its automation program with intelligent document processing and business orchestration

Connecting AI models and document workflows across a national mobility network

As the business grew, three different operational pressures converged. SOCAR receives traffic-fine notices, a large proportion of them on paper, in 287 different formats issued by 243 municipalities, with field positions and even field labels varying between forms. People manually keyed entries for 160 hours every month.

At the same time, SOCAR was operating 20 AI models for vehicle inspection, fleet operations, and maintenance, each running on its own with no view of cross-team work or handoff times. And as AI accelerated development velocity inside the company, manual regression testing, at five days per cycle, became the bottleneck holding releases back.

To handle the fine-notice problem, SOCAR deployed UiPath IXP (Intelligent Xtraction and Processing), an intelligent document processing (IDP) capability. From 287 formats, the team selected 32 representative types, built a labeled dataset from roughly 1,000 real notices, and stood up the workflow in about six weeks. IXP reads the text, classifies each notice by format, and pulls fields like municipality name, vehicle plate, violation time, payment deadline, and amount. Lower-confidence extractions route to a human reviewer, and those corrections feed back into model training. SOCAR runs the whole flow as an active-learning loop of inference, selection, review, and fine-tuning, so accuracy improves with use. Today, the extraction model runs at 95% accuracy, and missed entries that used to slip through manual handling are now caught automatically.

During a proof of concept project using UiPath Maestro™, SOCAR connected AI models and operational teams within a single workflow. The project demonstrated the potential to improve visibility, automate SLA tracking, and streamline cross-team coordination. Maestro opens a case and runs three AI models in parallel: exterior photo analysis, vehicle status, and tire wear. Each result is compared against historical data. If an anomaly surfaces, the workflow routes the case to the Incident Planning team through Jira and Slack, holds at a human-in-the-loop checkpoint, and resumes automated follow-up once a person confirms. Acknowledgment time and per-step service level agreements (SLAs) are now measured automatically, and cross-team handoff time is visible to operations leadership for the first time.

To keep validation pace with AI-accelerated development, SOCAR uses UiPath Test Manager in Test Cloud alongside accessibility-based screen structuring. Test cases are generated from structured screen definitions instead of being rewritten scenario by scenario, and Test Manager runs them on a continuous regression cadence. The work is still maturing, and reliable auto-generation of test cases is the hardest piece. Still, SOCAR has started with the SOCAR app and is expanding from there.

With UiPath, SOCAR now saves 96,400 hours of manual work each year. Its services operate at 99.98% availability, with incident resolution 97% faster than target. Active automation projects have grown from 7 to 59 across four years and now run in seven divisions company-wide.

Toward an AI-native mobility operation

SOCAR's automation capability has grown in stages. The program started in 2022 with software robots handling rule-based tasks. By 2024, the team had consolidated 20 AI models that had been running separately across different teams onto a single platform for deployment and monitoring at scale. Today, as the three cases above show, SOCAR is in the agentic phase: AI systems that take action under human oversight rather than only analyzing or recommending.

In 2026, the company declared the next step: becoming AI-native. Rather than bolting more AI tools onto existing processes, SOCAR is redesigning how it runs services and operations around AI agents as a core building block. The long-term goal is autonomous driving, moving from today's driver-assist systems to fully self-driving robotaxis, with the whole company aligned behind that direction.

The real ROI of automation, for me, is making possible what wasn’t possible before. The system picked up work the team had outgrown, and that freed people to move to review, judgment, and designing new services.

Junyeng Kim · Automation Innovation Team Lead at SOCAR

Outcomes

• End-to-end document automation: UiPath IXP processes the full range of traffic-fine formats SOCAR receives, with an active-learning loop that improves accuracy as people validate exceptions.

• Orchestrated AI workflows: A proof of concept using UiPath Maestro demonstrated how disconnected AI models and four operational teams could be coordinated within a single-tracked workflow, providing visibility into cross-team handoffs and SLA performance.

• Continuous test automation: Test Manager in Test Cloud runs auto-generated regression tests on a continuous cadence, reducing the manual QA bottleneck on AI-accelerated development.