ATM Cash Proposal System
ATM cash upload – Automation system based on AI / ML algorithms.
The system was developed in collaboration with a company of the Elfi Systems group.
Solution for ATM cash uploading
The most common systems for the daily generation of cash upload plans for ATMs are based on the following assumptions:
- To define the load amount, they use a simple static algorithm based solely on the value of the previous stock
- They apply the same algorithm without distinction to all ATMs
- They require the proposals to be approved by the operators, with possibility of manual correction based on the operator’s experience accumulated over time
Disadvantages of traditional systems
The systems normally in use are based on simple algorithms:
- They are not able to customise the proposal on the single specific ATM
- There is no “knowledge” coming from manual corrections applied by operators over time
- They do not allow the optimization of costs, due to the transport and insurance on cash stocks in the ATM
- They are not able to dynamically adapt to the evolution over time, upon the different needs of the individual specific ATM
- They have an important dependence on the manual intervention of the operator
- They require considerable effort in terms of time and attention of operators who have in charge the task of verifying every single automatic proposal
The proposed solution
Our innovative system dedicated to generate upload proposals for ATMs is characterized by the usage of a decision-making algorithm, based on artificial intelligence technologies, combined with sophisticated mathematical-statistical models.
The solution involves the use of different data sources, including:
- historical data on stocks and frequency of upload
- manual corrections made by operators (both historical and current)
- peculiar characteristics of each ATM (e.g. model, capacity, area)
- costs associated with the ATM (cash transport, storage insurance)
- manual training of the system for each ATM (operator feedback on the detected performance)
Due to utilization of artificial intelligence, the new system is able to process the acquired data and “learn” from them continuously, improving its effectiveness and reliability over time.
The result of using artificial intelligence is a natively dynamic and adaptive system, which offers a series of potentially specialized algorithms for each individual ATM which can respond in an increasingly customized way to withdrawal and deposit needs, and their evolution over time.
The system is not only able to respond to variations induced by known factors (e.g. holidays, seasons) but also implicitly by unknown events deriving from indirect circumstances (e.g. nearby new opening of an ATM point of another Bank, closing of important commercial activity nearby, etc.)
System Architecture
Benefits offered by the ATM Up-load Planning System
The new system allows to improve all the uploading operations, evaluating the proposal for every single ATM, while at the same time achieving the following results:
- Maximize the availability of money for each individual ATM based on the related withdrawal / deposit requests
- Minimize service trips in order to reduce transport costs
- Minimize ATM stock in order to reduce insurance costs
- Over time, the system “learns” and improves its effectiveness, threfrore reducing the dependencies on human intervention