AI Builds the Daily Payment Plan for Tens of Thousands of Suppliers
A large logistics company pays thousands of sub-suppliers every day. Deciding who gets paid today and who gets paid tomorrow is not a simple accounting task — it requires weighing cash flow, risk, and service quality all at once. In a consortium formed by Alışan Lojistik, Konya Technical University and Phexum under the TÜBİTAK 1711 AI Ecosystem Call, we built a cash-flow and financial-planning system that makes exactly this decision autonomous. We combined field experience, academic optimization power, and Phexum's data engineering. What the finance team used to do by hand every day was handed over to AI, gaining 200% to 300% in operational efficiency.
Featured Results
The system autonomously segments and scores thousands of suppliers and produces the optimal payment plan for each day. Once the manual payment planning and paperwork checks that finance and operations teams used to spend heavy effort on became autonomous, person-hour efficiency gained 200% to 300%. The system now autonomously builds the most optimal payment plan for each day without exceeding the budget limit. Early payments and faulty risk that could arise from manual processes were prevented. Due dates are balanced by AI recommendations, and cash flow became fully plannable. The project was registered by TÜBİTAK and the Ministry of Industry and Technology, and was awarded under the TÜBİTAK 1711 call.
Who the Client Is and the World They Live In
Alışan Lojistik A.Ş. is a large-scale corporate company operating in the logistics sector. The daily reality of a logistics company at this scale is managing a massive pool of thousands of sub-suppliers. Daily payment runs made without dynamic due-date planning and financial modeling create uncertainty in forward-looking cash flow. Even if today's payments are known, what remains in the till tomorrow becomes unpredictable. And the decision isn't purely financial: a supplier's past performance and service quality directly affect payment priority. Making this multi-dimensional decision by hand every day is both slow and error-prone. This was exactly the problem the consortium set out to solve.
Cross-Cutting Concepts and Architecture
The shared goal running through this entire system is paying the right supplier at the right moment. The first pillar of that is splitting thousands of suppliers into meaningful groups — technically, clustering. We did this with hierarchical clustering, K-means, and model-based algorithms, and validated result quality with scientific metrics like Dunn and Silhouette. On top of that, we brought in AHP (Analytic Hierarchy Process), a multi-criteria decision-making method. We scored suppliers not just on financial data but on past performance and service quality too — so payment priority now rests on the supplier's whole picture, not a single number.
The second core issue is speed at scale. Scoring thousands of suppliers every day would take hours with the wrong tool. That's why we trained machine-learning models based on XGBoost and neural networks to cut processing cost — so the entire pool can be scored autonomously, cheaply, and in reasonable time. The final layer holds the most critical decision. We modeled the daily payment budget's upper limit as the classic Knapsack Problem from computer science. We solved this problem — notoriously hard to solve in polynomial time — with nature-inspired metaheuristic algorithms (Ant Colony and Discrete Artificial Bee Colony), and integrated the result into the enterprise ERP infrastructure. This way, the most optimal payment plan is produced autonomously every day without exceeding the budget limit.
How the Collaboration Began
This project was born from three different strengths coming together. Alışan Lojistik brought the field's reality and its data. Konya Technical University added academic depth in optimization. Phexum took on end-to-end software and data engineering, from requirements analysis to data pipelines and algorithms. The TÜBİTAK 1711 AI Ecosystem Call gave this consortium both its framework and its support. This structure became a concrete example of bringing data-driven decision mechanisms into the logistics sector.
How the Story Was Built
We built the system with a hybrid approach that, unlike its counterparts in the literature, combines multiple AI and optimization architectures at once. As Phexum, we started with requirements analysis and built the solution as a four-stage data pipeline. First, we segmented the operational data of thousands of suppliers. Then we set up multi-criteria scoring. Next, we trained predictive models to run that scoring fast at scale. Finally, we modeled the daily payment budget as an optimization problem and produced the most profitable plan for each day. Each stage takes the previous one's output as input, and together they form a single autonomous decision chain.
The right supplier, paid at the right moment — decided autonomously, every single day.
Building Blocks of the Solution
The solution consists of four connected layers. The supplier segmentation and clustering layer splits thousands of suppliers into scientifically validated groups based on their operational data. The AHP and scoring layer performs autonomous scoring by evaluating financial data, past performance, and service quality together. The predictive models layer scores the entire pool cheaply and quickly with XGBoost and neural networks. The metaheuristic optimization layer solves the daily budget as a knapsack problem and integrates it into the ERP, producing the most optimal payment plan for each day.
The Impact We Created
The system made the payment plan and paperwork checks the finance team used to build manually every day fully autonomous. Person-hour efficiency gained 200% to 300%. Early-payment and operational-error risks were prevented, and cash flow became predictable. Combining academic depth with industrial engineering, the project was recognized as one of AI's most concrete outcomes in finance. And the framework and algorithms built here turned into a highly commercializable B2B FinTech product, adaptable to any domestic or international logistics company working with a large supplier base.