Predictive models that support managers in growing their dream teams and help employees advance their careers.

Technologies we used:

Course recommendation system

Course recommendation system for one of the world’s biggest international financial institutions

Employee retention is quite a challenge regardless of the industry. Managers have their goals regarding their teams but the team members also have their goals – they want to develop their skills and grow professionally. The goal is to align these ambitions, letting the managers meet their objectives and the team members advance their careers, either to a higher level of their role or to another role. The task is not easy, however, as it requires a lot of decisions to be taken both by the employee and the manager. 

When the Client approached us, the process of choosing courses for their employees was conducted manually. It required a lot of decisions to be made both by the employee and by the manager of a chosen department. Their idea to solve that problem was to implement Artificial Intelligence. The Client wanted to build an AI-based platform that would recommend the right courses to support the development of their employees and make sure that it aligns with the company’s goals.

The project’s goal was to streamline that decision-making process by creating an AI-based platform that would suggest the right courses by analyzing the requirements, career advancement opportunities, and historical data.  Simply put, we wanted predictive models to help managers grow their dream teams and to help employees advance their careers.

Challenges

1. Verifying the assumptions as soon as possible

2. Meeting the needs of different types of users

3. Understanding the current process of assigning the courses

Read more

Solutions

1. Running Neoteric AI Sprint at the beginning of the project

2. Building predictive models fed with historical data

3. All-round approach to the project, combining skills from different areas

Read more

Results

1. Proof of Concept, ready to test by the end users, built in less than 3 months

2. Hours saved by streamlining the process that previously required a lot of decisions to be made both by the employee and by the manager of a chosen department

Read more

Challenges

1. Verifying the assumptions as soon as possible

Implementing AI can be a long and costly process. Before spending years on building a complex solution, it is always worth to validate the assumptions with a Proof of Concept. The challenge is to find the right area where it is possible to quickly prove the value of such a solution.

2. Meeting the needs of different types of users

While it was crucial for the employees to get good training recommendations, at the same time it was important for the managers to make sure that their teams have the right skill set and that the courses help them get the missing skills. Also, it was important to consider that the employees may want to advance not only to a higher level of their current roles but also to different positions.

3. Understanding the current process of assigning the courses

Every corporate process has its business logic and it’s important to understand it as soon as possible as it helps align the product with that process and make sure that no part of that process is missing.

Solutions

1. Neoteric AI Sprint

The Neoteric AI Sprint consists of 2-day workshops with our team, and a Proof of Concept of a model that is chosen during the workshops. Its goal is to identify the right path to focus on in order to show the value of implemented models as quickly as possible. The Proof of Concept, which is the final product of the Sprint, is a real-life model trained on the historical data, helping the specific case which can be shown to the users as well as to high managers of a company or investors.

2. Building predictive models

Fed with historical data showing which courses were taken by different employees and how they impacted their careers, basing on similarities between the courses and different users, our predictive models are able to suggest the right courses to support career advancement within the structures of the institution. Depending on the goals of the employee and the assessment of skills needed in every team, the models are able to suggest the right courses that make the training process more effective.

3. All-round approach

Even though the core of the project was about building and training predictive models that would recommend courses to employees, creating a good Proof of Concept required combining skills from different areas as well. In order to make sure that our Proof of Concept is scalable, we used serverless architecture for its backend part. As we wanted to provide the Client with a usable platform, we involved a UI designer and a frontend developer to take care of the visual part of the project.

Results

1. PoC built in less than 3 months

From the very beginning, it was important to validate the idea of implementing AI as quickly as possible. After the AI Sprint workshops, we agreed on the problem that we wanted to solve with predictive models and on the project scope. The next step was to deliver a product that would prove to be a solution to that problem that is worth to invest in.

The Proof of Concept that we’ve built using the data provided by the Client has met their expectations, being able to suggest the right course recommendation for their employees. As the models can learn with every new data record they get, their suggestions are more and more accurate every day.

2. Hours saved by streamlining the process

The platform was able to streamline the process that previously required a lot of decisions to be made both by the employee and by the manager of a chosen department. Now, the platform is able to suggest the right courses that would support managers in growing their dream teams and help employees advance their careers. The managers are able to see their teams’ development in a visual way and quickly assess what skills are missing, while employees get course suggestions as a clear list of recommendations – just like the way they get movies recommendations on Netflix.

Technologies used

Python

Python

Angular

Angular

AWS

AWS

Azure

Azure