We created a small piece of web-based middleware that pulls in data from both RT and FreshBooks and then generates its own new data based on the input it receives. WCI accesses the new data through a rights-based interface, and the new functionality gives WCI extra capabilities in both RT and FreshBooks while limiting access to a small group of end users.

Getting the data in was one thing, but making sure we had it right was something else altogether. Although WCI’s engineers and clients change over time, certain patterns began to emerge the more we looked at things.

We believed these patterns held the key to verifying what went into the invoices, so we applied some basic machine learning algorithms to the data. And, our models have been very helpful at identifying possible problems before they reach live invoices.

Finally, to ensure that everything runs smoothly, we created monitors and alerts to tell WCI if there's a problem generating an invoice or if there's data that needs to be manually reviewed. Additionally, we gave WCI the ability to simulate an invoice run for any date, helping them better manage income flow and billing cycles.