Skip to content
Dramatically Reducing Customer Support Load with an In-App Conversational AI Solution

Banking, Financial Services & Insurance

Dramatically Reducing Customer Support Load with
an In-App Conversational
AI Solution

Dramatically Reducing Customer Support Load with
an In-App Conversational
AI Solution

Success Story

Our client, a FinTech company, sought to increase user engagement and retention rates by boosting their customer support process through AI. The Scalefocus team helped them develop a custom AI solution handling thousands of chat sessions daily by employing a combination of cutting-edge OpenAI large language models (LLMs) and smaller fine-tuned models. The solution communicates within the context of complex financial regulations and currently serves over 65% of all chat sessions, freeing up agents’ time for creative solutions to non-repetitive tasks.  

Excellent handling of increased call volumes

30%+ of load reduction for support teams

65%+ of sessions served by the solution

30+ languages with unlimited extensibility

The Client

Our client is a fast-growing finance and investment company democratizing the financial and stock markets for people around the globe. A FinTech at its core, the company offers intelligent and intuitive mobile applications that customers use to access financial and trading services in the palm of their hand, wherever they are. More than 2 million users across the globe take advantage of the client’s impeccable financial platform, which is regulated in the UK, Europe, and Australia. 

The Challenge

The client wanted to provide an even better customer experience to a fast-growing customer base through an automated solution for their in-house support center. Our team set out to build a custom AI solution as conversational AI has proven to be a game-changer across various industries in saving time and improving results. We had to fully integrate the solution with the client’s customer support systems and with the satisfaction rating system used for improving customer care.

The goal was to free up support agents’ time for creative solutions to non-repetitive challenges while enabling them to manage and expand the knowledge base of the solution.

The Solution

The solution we designed together with the customer has a pluggable architecture using proven technology, allowing the underlying models to be easily swapped for third-party services or on-premises solutions. Currently, it employs a combination of cutting-edge OpenAI large language models, giving the best results in conversational customer care, and smaller fine-tuned models (currently Llama 2 that can be switched due to the pluggable architecture), which are employed in various classification tasks – e.g., the bot can detect intent for a conversation with an actual agent and redirect customers seamlessly.

The solution skillfully communicates within the context of complex financial regulations—it keeps personal data confidential, uses only anonymized data, never gives investment advice, uses disclaimers, and provides transparent information on risk. Since maintainability is of the utmost importance in the whole ownership lifecycle, we created a testing evaluation suite monitoring for performance and veracity degradations. It measures the quality of all answers and includes a human assessment step optimized for evaluations at scale. 

Great AI solutions are always built on quality data, therefore the knowledge base of the AI bot we created is stored in a vector database and managed through custom UI. In this way the staff can easily and securely manage it by adding new articles of knowledge and operating with existing data. Empowering the support teams to manage the knowledge base is especially useful in crisis management (e.g., when there is a service outage, and many customers are contacting support with similar queries). It also enables the support teams to automate repetitive work with the urgency needed. 

To provide the support teams with relevant data and widen the automation impact of the solution, we integrated applications throughout the client’s system landscape. The data transformations on the knowledge base are optimized for retrieval and usage in large language models (LLMs). The second version of the solution will have secure private access to customer account information maximizing customer experience and enabling self-service support.

The Results 

Our team delivered a fully integrated custom AI solution that is now operating across the client’s in-house customer care system. It acts as an entry point for all user interactions and handles thousands of end-to-end chat sessions per day. The chatbot technology automates tasks and significantly reduces the volume of support tickets, enabling specialized individuals to focus on more complex issues and provide a higher quality of service. 

We were able to achieve: 

  • 65%+ of all chat sessions currently are being served by the chat bot 
  • 30%+ reduction of load on the customer service teams in normal operations 
  • 70%+ reduction in load on the customer service teams in surge events 
  • 30+ languages supported and unlimited language extensibility 
  • Elimination of the need for overprovisioning in customer care 
  • Seamless AI-driven escalation to a customer service agent

Technologies

ChatGPT
API integrations
Zendesk integrations
Llama 2

Our Work

We have a global client base that includes Fortune 500 companies, innovative startups and industry leaders in Information Technology, E-Commerce, Insurance, Healthcare, Finance and Energy & Utilities.

Explore all

Ready to scale and meet the technology challenges of tomorrow?

Contact us