The purpose of the features section of this development portal is to give you an understanding of what you can build with the AzoraOne API. The examples given are meant to be understandable whether you are a business person, a product person or a developer. For the technical nitty gritty - head over to Documentation!
Below you will learn about basics of AzoraOne, and this introduction will help you to understand the other pages better.
The AzoraOne basics
The AzoraOne API is a niche product, meant to solve one hard-to-solve problem; the automation of data entry from financial documents. The typical user of the API is a company providing an accounting software to SMEs and/or accountants. It can also be used for smartphone apps such as receipt apps, or by partners/companies that build integrations to an ERP system. The company that integrates AzoraOne does that so that the value for the end user appears embedded in the user interface of the ERP/accounting software/smartphone app that person is already working in. The result (automation of data entry) is instantaneous and presented in a manner the software company has chosen, for example in the form of pre-populated fields in the user interface for accounts payable.
Example below from Swedish accounting software Briljant. As different types of financial documents are uploaded to the software, AzoraOne is working in the background. The presentation of results in the Briljant UI are triggered by the click of the "Enter" button on the keyboard - and results are presented before the eyes of the end user within a few hundred milliseconds. You can learn more about this integration in this case study on LinkedIn.
The self-learning process
At the core, AzoraOne consists of an algorithm capable of continuously learning by observing financial documents and the data entry associated to those documents. There are no back-office humans keying in data - the AzoraOne API strictly consists of algorithmic automation.
The knowledge is, in the base case, isolated per company. It is continuous, and it is delivered in real-time.
Let us look at a scenario where AzoraOne has been integrated into an accounting software, to explain the basics:
The end user uploads a document into the accounting software
The accounting software automatically uploads the document to AzoraOne
AzoraOne uses prior accumulated knowledge - and the document is sorted into the correct type
AzoraOne uses prior accumulated knowledge - and extracts base and categorization data
The type, base and categorization data can now be presented in the accounting software
The end user either accepts the data or makes appropriate changes
The accepted/changed data is sent to AzoraOne and adds to the continuous learning process
Upon building the integration, you control the different steps in the communication with the AzoraOne API. A well built integration will lead to a real-time experience for your end user.
Learn more about this process in technical terms in the Documentation section!
Company specific knowledge
The knowledge of how the end user wants particular values structured in the end point system is accumulated and (in the base case) isolated in a company level. This means that the accounting software that has integrated AzoraOne gains the capability to automate categorization of costs in a very precise manner; in accordance with how each company chooses to categorize costs.
The knowledge can also be inherited between companies in a controlled manner using the resource Progenitors.
Super rapid delivery
As the integration partner, you can choose when to execute certain actions in the communication with AzoraOne in a way that leads to blink-of-an-eye delivery for your end user. For example you could choose to upload a document as soon as it is uploaded/emailed into the accounting software; AzoraOne extracts the data and once the end user enters the user interface, you can present the results in real-time, in the matter of milliseconds.
AzoraOne differs from common Machine Learning approaches in three important ways:
First of, AzoraOne is not dependent on huge amounts of data or training data sets. You could integrate AzoraOne into your accounting software, upload a very first document, bookkeep it - and see results on the second document. Of course, to get good results you would need more than that, but not that much. On a company level, you could think of it as "a couple of invoices per vendor" to be in the right ballpark.
Secondly, the learning is live and continuous in AzoraOne. Meaning: There is no retraining of clunky machine learning models, no updates of data sets, no cleaning of data etc. Once integrated, AzoraOne just keeps learning and the maintenance for you is close to none.
Thirdly, the knowledge of AzoraOne is not a black box - the behavior of the algorithm can be tracked down to a document level. This means great control over the delivery of the algorithm.
Maintenance for you, the integration partner
Well, hardly any. There is no retraining of machine learning models. No mapping, templating and so on. Once the integration is built you will end up with a solution that is rather hands off - and you can focus your developer team's efforts on other issues at hand.
We at the Arkimera Robotics team continuously add functionality to the API, so it will naturally be up to you to prioritize taking advantage of such updates. Since we have created the technology behind AzoraOne from the ground up, we are not reliant on the quality of other companies' Machine Learning models, templates and so on. For you, our partner, this also means a great opportunity to be part of the development of the API.
AzoraOne Sandbox API