Artjoker
Case Studies
AI solutions
AI OPERATIONS FOR 85 FUEL STATIONS

From dipstick checks to a self-running network:
AI operations for 85 fuel stations

AI AGENTIC DEVELOPMENT

Project Overview

A regional fuel station network with 85 locations across Ukraine was facing growing operational complexity: fuel inventory management, shift scheduling, loyalty program administration, market-driven price adjustments, and regulatory reporting were all handled through disconnected systems and manual processes.

The client's IT landscape included a legacy station management system, a separate loyalty platform, and Excel-based reporting for regulatory filings. Artjoker was tasked with building an AI agent layer that would unify operational data, automate routine decisions, and give network managers real-time visibility.

8–12 MIN

Price adjustment cycle — down from 3–4 hours
↓ 80%

Fewer unplanned outages from fuel shortages
85

Stations unified under one AI operations layer

The Challenge

3–4 hour approval chains

for fuel price adjustments — the network lost margin on fast-moving spot market shifts.
12–15 outages a month

caused by manual dipstick measurement errors across the network.
~$95K/year in loyalty fraud

card sharing and fictitious transactions caught only during monthly manual audits.
40+ hours a week

spent by the HR manager scheduling shifts across 85 stations and 340+ employees.
30% of deadlines missed

regulatory reporting relied on manual data aggregation from all 85 stations.
No centralized detection

equipment failures, cashier errors, and fuel losses were caught reactively.

The Solution

Artjoker built a unified agent layer connecting all station systems to an intelligent network control center — replacing six disconnected tools with one always-on operations hub.

Agent 01
Fuel Inventory Monitoring Agent

Integrates with ATG sensors across all stations in real time. Applies anomaly detection to flag abnormal consumption from leaks or theft. Forecasts tank levels 24–72 hours ahead and autonomously initiates delivery orders.
Agent 02
Dynamic Pricing Agent

Monitors spot prices (Platts, ICIS), competitor prices via a field-agent app and web monitoring, and target margins. Proposes adjustments within 8–12 minutes of market moves and pushes approved prices to POS instantly.
Agent 03
Loyalty & Anti-Fraud Agent

Continuously analyzes transaction patterns to detect card sharing, velocity anomalies, cashier collusion signals, and fictitious refueling. Suspicious accounts are auto-blocked pending human review.
Agent 04
Shift Scheduling Agent

Uses historical traffic data and event calendars to generate optimized shift schedules. Automatically processes shift swap requests and checks compliance with labor law.
Agent 05
Compliance Reporting Agent

Aggregates quality-control data from all 85 stations and auto-generates regulatory reports in DSTU formats, with draft reports and reminders 5 days before deadlines.
System
Unified Operations Dashboard

All five agents feed analytics into a single real-time dashboard, with mobile access for regional managers on the move.

Tech Stack

AI / LLM

Claude Sonnet (orchestration, reporting), time-series anomaly models
Agent Framework

Proprietary multi-agent orchestration + LangGraph
Data Sources

ATG sensors (Veeder-Root API), POS systems, Platts/ICIS API, DSTU database
Price Updates

Proprietary POS integration (real-time price sync)
Competitor Monitoring

Field agent mobile app + automated web scraping
Loyalty

API integration with the loyalty platform
Shift Scheduling

Shift management module + mobile app (React Native)
Dashboard

React + WebSocket real-time updates
Infrastructure

Azure AKS, TimescaleDB, Kafka

Results

Metric	Before	After	Change
Price adjustment cycle	3–4 hrs	8–12 min	↓ 95%
Unplanned outages from fuel shortages	12–15/mo	2–3/mo	↓ 80%
Loyalty program fraud losses	~$95K/yr	~$18K/yr	↓ 81%
HR time on shift scheduling	40+ hrs/wk	6 hrs/wk	↓ 85%
On-time regulatory reporting	70%	100%	↑ 30 pp
Fuel level variance (ATG vs POS)	1.8%	0.4%	↓ 78%
Control center operator headcount	12	7	↓ 42%

We went from reacting to problems the next morning to catching them the moment they happen — or before they happen at all. The tank monitoring and anti-fraud agents alone paid back the project in 5 months.

— COO, Client

Key Expertise

Real-time IoT integration

A network of ATG sensors across 85 distributed locations.
Competitive intelligence automation

Combining field data and web monitoring into a single pricing agent.
Multi-system orchestration

6 previously siloed operational systems unified under one AI layer.
Proactive anomaly detection

Fuel inventory, loyalty transactions, and equipment condition.
Compliance automation

Reduced regulatory risk without manual report preparation.