Big Retail Bank Gets a Competitive Edge with Conversational AI


One of the biggest banks* in Bulgaria in terms of assets, with a broad branch network and considerable client base, has implemented a wide range of modern alternative banking channels. The almost three decades of its successful business operations have positioned The Bank as a benchmark for innovation, shaping the country’s banking trends in recent years and creating award-winning products.

The bank has decided to catch up with the growing trend of using sophisticated tools such as Big Data management, digitized analytics, and artificial intelligence, which are becoming the key to successful financial management and are transforming the banking world. As a result of this innovative vision, in 2020, The Bank started a large-scale project of implementation of a conversational AI based on Sirma’s chatbot platform (available branded as Melinda or as a white label). The main goal was to improve the Customer Service Center’s key performance metrics and enable the financial advisers to offer more relevant and competitive services.

The project scope

The main project requirements were that the solution should use natural language understanding (NLU) and a self-learning chatbot that answers simultaneously to an unlimited number of customers. It was expected to support conversational interactions in both Bulgarian and English over multiple touchpoints, such as Web, Mobile apps, and Messaging apps - Viber, Facebook Messenger, and Skype. The solution should also have integrated speech-to-text and text-to-speech functionality.

The implementation of Conversational AI has been planned in two phases

Phase 1 - implementing an Intelligent Chatbot, servicing existing and new customers through different channels, with no need for customer identification. The chatbot should be able to interact with the customer over the following topics:

  • ATM/Branch locator
  • Tariff (simplified)
  • New product application
  • Frequently asked questions (FAQ) including but not limited to FX rates, General product information, Lost/stolen debit/credit cards, E-banking / Mobile banking support, News, and others (e.g., BIC, SWIFT, etc.)

In addition to the functionalities implemented in Phase 1, the next project step includes the integration of Conversational AI with the bank’s systems over ESB (Phase 2).

The final result that the client was looking for was a multifaceted solution that supports: active banking operations like delivery of personalized banking data: accounts, and transactions related requests on information; balance checks; onboarding new customers; transaction processing as an alternative or add-on to the traditional digital channels; utility payments initiation and schedule; PFM and Robo-investment advisory; cards blocking; fraud management and booking an appointment with bank specialist. Customer identification has to comply with the PSD2 requirements.

The Goals

  1. The ultimate goal of the Bank was to implement an innovative and easy-to-use solution that improves the existing customer service and attracts new customers.
  2. The Bank wanted to provide its customers with a quick and easy way to answer customer questions and offer appropriate solutions to meet the needs of its customers through human-like speech and helpful conversation.
  3. The Bank was looking for an Artificial Intelligence (AI) and Natural Language Understanding (NLU)-based solution that would automate the entire customer support process - receiving questions, understanding the customer’s intent, and providing the appropriate answers.

The Challenge

The two teams’ main challenge was to combine the best of business, technological and human intelligence to solve the fundamental problems mentioned above. The solution should connect cutting-edge AI and machine-learning technology with the available requirements and put the banking experts out of their comfort zone while designing and developing a virtual assistant for various communication channels.

Another challenge was to cover and keep seamless communication between the bot and the customer. Finally, the project execution and implementation were performed remotely due to the COVID-19 lockdown.

Why Melinda was chosen as the platform foundation to build the new Conversational AI

Sirma is a company that has 20+ years of experience in applying various advanced AI technologies like NLP, text mining, and many others; thus, it was the preferred choice for developing this intelligent assistant. The next-generation bot, based on our platform Melinda, was empowered by artificial intelligence, especially NLP/NLU and machine learning (ML) technologies that mimic human conversations and provide automated proactive interaction with customers.

The financial institution has selected Sirma as a technology partner that would help achieve the following outcomes:

  • Decreasing the Call Customer Service Center and Branch network’s workload;
  • Call Customer Service Center’s staff optimization;
  • Significantly improved key performance metrics such as abandoned rate calls and response time;
  • Fewer Agents’ time spent on routine questions and more time on sales activities;
  • Boost Self-service sales and cross-sales based on a marketing campaign or advice while using the bot;
  • Better 24/7 customer experience with decreased rates of missed or waiting calls;
  • More happy and loyal customers.

How soon can Conversational AI become a trusted partner to financial advisors?

Conversational AI can help each financial organization fill the communication gaps between their business and audience, set up additional communication channels, and actively offer products and services. Even more, many new services, such as providing financial advice, detecting fraudulent actions, and facilitating back-office optimization, can be considered new banking use cases. The potential of the technology can be unleashed by providing personalized financial recommendations or assisting with financial planning tasks. It’s important to note that while conversational AI can provide helpful information and assistance, it is not a complete substitute for the advice of a trained financial professional.

The Results

Phase 1 of the project finished in the planned timeframe, and the bank implemented the chatbot assistant in multiple channels.


  • We designed and developed a powerful AI-driven chatbot platform, which helps the bank answer its customers’ questions in the blink of an eye, identify potential new customers and opportunities, and transform how it works with them.
  • Both teams succeeded, working remotely and independently (the COVID pandemic didn’t allow teams to have even one person-to-person meeting).
  • Independent research by the bank confirmed that natural conversation with customers is very well accepted. This communication differs in their options compared to the front-end staff, so financial institutions should consider covering them.
  • The bank has not only revolutionized its interactions with customers but has also opened up new opportunities for sales.
  • What our two organizations fulfilled (beyond the goal at hand) is a proof of concept for other banks and technology companies. The traditional on-site banking systems delivery can be performed remotely without needing the physical presence of all teams – from the bank and the technology vendor sides.

As the project is ongoing, the forthcoming Phase 2 will include an upgrade to the UI and UX of the solution with new active banking workflows, along with the integration of Conversational AI with the bank’s systems over ESB. The latest enhancements will define a different added value to this conversational channel. They will make it an alternative to the standard online banking channels like the Internet and mobile banking, phone banking, digital wallets, etc.

*According to NDA, we cannot disclose the client’s name.

Transform your business