🧪 Beta tests of Alphamoon's automation platform are open. Sign up and process invoices for three months for free.
28 Sep 2019

How automating a CRM increased sales by 10% for a European bank

Our client is the Polish branch of multinational commercial bank and financial services company operating in Europe, North and South America, and continental Asia. As the fifth largest bank in Europe, our client faces tough competition on the market. To increase its efficiency in reaching customers, the company looking to launch a digital transformation project in its sales department.

Extracting information from a massive dataset

Our client approached us with a specific problem. The company had a database containing more than 100 thousand notes sales reps made during conversations with clients. Naturally, these notes offered a wealth of contextual information that would help the bank make more relevant offers. 

However, manual processing of such a vast data volume was next to impossible. Our client was aware that artificial intelligence (AI) solutions could deal with such a challenge. That’s why the bank’s goal was creating a system that would process all the data contained in notes automatically: an AI-powered note analysis engine.

But that’s not everything. That system would also add insights from data into the company’s CRM platform to help sales representatives in their daily work. For example, the solution would support sales reps in choosing the best-suited prospects to contact with new offers or check which offers are the best match for a specific customer. 

Outsourcing done right

We carried out a series of meetings with the client to better understand the unique requirements of the project and, more broadly, of the organization. 

Our team of two Machine Learning experts got down to work within two weeks after setting its requirements. The two-person team was complemented by a senior Machine Learning developer who supervised the work and was responsible for choosing the technology stack and Machine Learning models. Following the agile methodology, our team has completed the project during the following eight weeks.

What about the client’s side? The bank appointed a product owner who oversaw the project, ensuring that its development was aligned with the organization’s business requirements. 

Here’s how we built a note analysis engine

Our team of Machine Learning experts developed a library with an API that expanded the CRM system with new functionality with the help of Python. Using Deep Learning and Natural Language Processing (NLP) techniques, the team implemented the following modules:

  1. BPE (Byte Pair Encoding) – the first step was building a dictionary of tokens. In our case, it was a dictionary based on the document database shared by our client. We used a coding technique that ensured our solution could deal with the errors and typos that appeared in notes frequently, as well as vocabulary and shortcuts specific to the bank.
  2. GloVE (Global Vectors for Word Representation) – we used this technique to build embeddings (continuous representations) for tokens that were created during the development of the dictionary. Another goal of this stage was learning the semantic similarity between the words appearing in the notes. In particular, our solution needed to handle the jargon and abbreviations used by sales reps working in the financial services industry.
  3. Bi-LSTM model (Bidirectional Long Short-Term Memory) + MLP (Multilayer Perceptron) – we implemented a Deep Learning model that was trained to predict the probability of customer purchases on the basis of notes represented by embedding chains. Based on the predicted probabilities, customers were sorted and the most profitable group was chosen to be followed up.

Results of our collaboration

Once our team implemented the solution, the bank could finally unlock the business insights contained in their massive data set. That lead to c. 10% increase in sales, helping our client’s sales reps become more productive in their daily activities. They could now approach customers with the full context of their previous interactions with the bank. 

The success of this project marked the beginning of a fruitful long-term collaboration that has lasted until today. We’re happy that our experts get to support a significant European financial institution in optimizing their operational efficiency and driving growth.

If you’re looking for a team of expert software engineers with AI skills, get in touch with us. We know how to create solutions that help organizations make the most of their data.

Less paperwork. More time for business.

Learn more