Hubble is a Dutch company offering state-of-the-art vehicle fleet management systems. Their products were designed ‘by the fleet managers for the fleet managers’, with the goal of saving time, money and workload for those in charge of a company’s car park. We’re proud to say that what started quite casually as a proof of concept during one of our Flock. Days grew into a full-blown cooperation with Hubble. The goal was to automate manual processes, free the manpower to do other tasks and, last but not least, scale-up the operations in the future.
Soon enough we were working on our first tool for the company, building an application that coordinated a large number of scrapers to extract reports from the websites of lease companies. We wanted to automatically retrieve vital chunks of information, such as number plates, user, billing, CO2 output etc. and export them into the Hubble environment. The idea was to do away with tedious and repetitive tasks performed by humans and replace them with an automated tool which would make the data available in a timely manner.
The next issue to be addressed had to do with invoicing. As a matter of fact, some larger clients of Hubble invested a lot of time into receiving and registering the invoices manually, employing up to one FTE for the task. Needless to say, that person’s job wasn’t the most fulfilling - and here’s where Flock. stepped in.
The main challenge was that the invoices come from various parties (e.g. car repair shops) in a wide variety of formats, with all the relevant pieces of information appearing in different places in the attached pdf files. Therefore, as a first step, we set up Mailgun to parse the incoming emails and identify the attachments. We then extracted the pdfs and sent them to an invoice document AI model. The latter would subsequently send them back in a structured way, with number plates, invoice amount and damage identification exported neatly to the Hubble fleet management software.
While the AI model proved to be reliable for the majority of the invoices, we needed a tool to manually check the ones we identified as unreliable by using various integrity checks. In order to achieve that, we built an interface which enabled manual modification of the data should, for example, a licence plate be missing or the totals not match. In the end, the whole array of data from all the different sources and formats found their way into a neat Hubble database, an improvement which meant Hubble manpower was freed to perform other tasks. More importantly even, the automation of manual processes will allow the company to move on with their scale-up ambitions in the future. At Flock. we’re immensely proud to have been part of this process.