The Making of Alix

Share This Post

A Discovery tool for intelligent automation

There is an increasing demand for Artificial Intelligence in the market today. The demand is ridiculously high to the point of becoming akin to expectation: once a new product or software is rolled out, the inevitable questions are immediately asked: Is it smart? How does it affect me?

Attempting to define intelligence

Intelligence is a concept that eludes us to this very day. Understanding it has haunted mankind for the better part of its existence, and today this curiosity has been refined and reshaped into a field of its own dubbed “Artificial Intelligence” (AI).

Natural Language Processing (NLP) is a subdomain of AI that combines intelligence and language benefiting from a multitude of applicable machine learning and deep learning techniques to help machines become “smarter” and manipulate language with ease as humans do. While NLP promises to be one of the leading frontiers of AI for the coming years, it is important to assess the following questions: Are machines ready to become smart? How smart are they capable of becoming? Are they going to be smart enough to help me?

Enter ALIX

As AI researchers, finding ways to bless inanimate objects with A form of intelligence is part of our work. And while AI remains a wild horse – unpredictable and difficult to tame – we proposed to evaluate how much of its power we could harness to breathe some cognizance into one of our products: ALIX.

Simply put, Alix was born with the purpose of identifying opportunities for automation. We, as pioneers in Natural Language Generation (NLG) and leaders in Global AI Software, decided this is a job for us. 

In order for ALIX to be able to make such a crucial business assessment, it needed to learn to read financial reports and cleverly recognize the portions that can be automated from existing data at Yseop.

ALIX meets AI

There are two leading intuitive ways of thinking about AI and knowledge transfer:

  1. Possession of large amounts of information (data) regarding a certain topic and falling back on this data to make informed decisions in an automated fashion.
  2. A rule-based system whereby the decision-making process is derived by going through a series of conditions and testing their validity. The conditions are structured in a business context. 

In the case of ALIX, one of the main challenges faced was generating a substantial corpus containing the information needed to build an accurate tool. Because of this, it was not possible to rely on large amounts of data for knowledge transfer. Specifically, to make ALIX capable of measuring the level of automation in a report, we needed it to look for special information in examples that have already been automated by Yseop’s platform. This effectively ruled out the data-driven knowledge approach, and we turned to the rule-based approach instead.

We benefit from the existence of a “rule bank” which can help it make informed decisions regarding a report. Working from the available numerous examples of automation, and aided by a thorough linguistic analysis, we successfully reverse-engineered the rules that define the markers to look for in a report that makes it a candidate for automation through Yseop.

The set of rules are complex and very deep, analyzing everything from sentence structure to types of data and verb/noun relationships.

ALIX learns these rules like a human and tests its understanding by trial and error. It is then capable of performing calculations based on the existence or non-existence of custom markers to generate an overall score of automation for a report. This effectively makes determining the automation level a quantifiable task and thus Evaluates automation opportunities.


Building ALIX was certainly a challenge, but it helped us get closer to answering the primal questions: Are machines ready to become smart? And how smart can they really become? Can I harness this technology?

There is a firm belief in the scientific community that intelligence does not come in one form, but rather in an umbrella of different kinds. ALIX is proof of this belief, as it demonstrates that it is not necessary to know everything to be able to make informed decisions. Rather, possessing the relevant information suffices to navigate specific use cases. This resonates deeply with the way we think as human beings, and we continue to prove that not having all the information in the world in our brain does not hinder us from having the right amount to make smart decisions.

Today ALIX continues to grow and learn as an eager (and popular) financial analyst and is quickly finding its feet in its new role.

Enough talk, see for yourself how much of your reports can be automated, we know you’ve been wondering.

Article written by Hanna Abi-Akl from Yseop

Scroll to Top