™ brings artificial intelligence models to the Business Administrative Operations (BAO) realm, relieving workers of tireless manual document handling and freeing them up for business-critical tasks.

3 documents which have been correctly verfied by StaxAI

Consider It Sorted

Manually managing Business Administrative Operations (BAO) can be frustrating for employees and expensive for your clients. was developed to automate and optimize BAO.

BAO is the process and operations surrounding the paperwork, documents, data, and information in your company, as well as the transactions within a business to keep it running.

Using™, organizations will reduce the amount of time spent handling and digitizing documents in all of the four main areas for document operations:

  • Intake
  • Distillation or Filtration
  • Search
  • Storage

The team’s vision is a world where businesses can reliably automate the entire document lifecycle, without expensive, error prone, manual tasks. From the moment a document is received all the way through permanent virtual storage, helps organizations of all sizes take control of their documents — freeing employees to focus on business-critical tasks.

Stax SortStax DiscoverStax Onboard

Building A Smarter Cloud

In 2019, completed an initial study with input from hospital administrators and early insiders. Their findings were that it costs a small 10-person organization around \$330 monthly for a single worker to file and sort documents, and this only took into account the initial sorting into mailboxes. This expense escalates quickly as more specialized workers, with much higher rates of pay, start interacting with documents upstream.

After their thorough validation, the team knew they could help make a major impact on BAO. The next question they needed to answer was, “How will we deliver this product?”.

They decided early on that they would need to build a scalable, distributed application that could take advantage of complex deployment patterns. The obvious choice for delivery of their application was containerization, but they needed to decide how they would handle the inherent technical debt associated with container orchestration and infrastructure management.

With these considerations in mind, the team decided to use the Cycle platform to standardize and automate their infrastructure and container management. This gave them more resources and time to focus on their core product.

“We saw in Cycle an opportunity to believe in the same message we were promoting. The platform itself would reduce the amount of manual work our team would have to do and at the same time keep our specialized talent away from tasks that lead away from our core offering. We were also excited to pass on the cost savings to our end-user”

- Team

Having the developers use Cycle gave management the ability to reallocate resources from a DevOps team to a team of AI specialists. With the ability to hire for these specific roles, the core product launch timeline was cut down dramatically and the team was able to pursue a much more robust feature list. By the time was launched the team had developed a highly specialized product in record time that was more affordable for the end-user.

Filling In The Gaps has been live and delivering solutions to clients for over a year now. One thing the team noticed was — when the training set for their stacks was very small (less than 5 documents per stack), the platform was less accurate.

Recently, Derek, one of the brilliant Solutions Developers for completed research on how they could improve performance on these smaller datasets. The research yielded creative pre-processing techniques that have greatly improved’s model performance. The pre-processing of the data has helped the existing models move from 67% accuracy from 5 documents to 94%.

“The update includes optimal class weighting for training, along with the use of TF-IDF (Term Frequency - Inverse Document Frequency) to vectorize (how you get a machine to understand words and language), and the use of a new intelligent over-sampling method that generates synthetic training data.”

- Team

As if that wasn’t enough, the team has also hit major breakthroughs in hand-writing recognition where the model can now read most English handwriting. Check out these amazing before and after shots showing the improvement.

Stax Handwriting 1Stax Handwriting 2

Needless to say, we are always inspired by the team, and all they continue to accomplish. To check out everything they are working on head over to their website.