AI is an incredible hype, and that often creates two camps. Some people are very cautious and mainly see AI as a hallucinating people-pleaser. Others believe AI will solve all the world’s problems.
At SiNube, we look at AI differently. We simply use it to solve customer questions in a more efficient way. Below are three examples of how we use AI in practice.
Has my invoice already been paid?
When suppliers contact the finance department, the question is almost always the same: “Has my invoice already been paid?” It is a simple question, but in practice, it can take quite a bit of manual work. An employee needs to read emails or take notes from a phone call, identify the supplier, check the invoice details, confirm that the invoice exists, look up the payment status and then provide a correct answer. With high volumes, this can quickly become time-consuming.
Can we automatically match all payments?
A finance employee may only need to manually match around 10% of payments, while all other payments are automatically matched to the correct document. Again, this is a simple task, but manual work is still needed to handle the recurring exceptions: missing or incorrect references, partial payments, combined payments, partially disputed invoices or amounts that do not fully match.
Checking overdue customers even faster
Our customers also ask if they can quickly see which customers have not yet paid their invoices. This information is already easy to access, but they want it even faster and on any device, including for people who do not work in the finance department.
So, could we use AI to answer payment questions, match payments and quickly find overdue invoices? Of course we can. We recently proved it by building a working prototype with a team of six during the Oracle Hackathon.
SiNube wins the Oracle Hackathon Benelux
During the Oracle Hackathon 2026, we showed what is possible with AI. Our use case was based on a familiar situation in accounts payable: an organisation receives a large number of supplier invoices every year and regularly receives reminders from suppliers asking about the status of their payment. Each question can easily take 10 to 20 minutes to process. That is time finance employees could use for more valuable tasks.
The solution we developed during the hackathon shows how AI can support this process. A supplier sends a reminder or question about an invoice. The communication is automatically read and understood. The AI agent checks the relevant information, verifies whether the invoice is known, validates the data and helps determine what status can be shared with the supplier.
Our hackathon team focused mainly on reliability. The solution performs checks for possible phishing attempts or incorrect information. It also uses the access rights and data that a finance employee would normally have access to, so information is not simply retrieved from anywhere or shared freely. Where needed, we also built in human checks. This human-in-the-loop approach ensures that the automation remains controlled and secure.
The result is an automatically generated response to the supplier with the payment status of the invoice. In a further version, we could add additional information, such as payment confirmations or an explanation when an invoice has not yet been processed or is blocked.
The value of an AI agent is clear in this use case. Suppliers get an answer faster. The finance team spends less time on repetitive questions. For the organisation, this is a step towards touchless finance: financial processes that run as automatically as possible, while keeping control and security in place.
The main lesson from the hackathon is that AI in finance does not have to be abstract. The value lies in very concrete, familiar processes: the same questions, the same emails, the same checks, every day. By intelligently automating these repetitive tasks, teams can work faster, provide better service to suppliers and give finance employees more time for more valuable work.