Quickfox Technologies

AI MVP of GenAI-Powered Chatbot:

The client operates the country’s largest 4G network, serving 17.1 million customers with high-quality, modern, and cost-effective services, delivering value to both customers and partners.

To further enhance their customer support, a custom AI-powered chatbot solution was developed using generative AI and advanced NLP capabilities. This intelligent chatbot seamlessly integrates with the company’s SMS API and centralized knowledge base, enabling real-time responses to customer queries. The system also automates repetitive tasks, significantly reducing the dependency on human agents.

Challenge:

The client had an existing chatbot with basic NLP capabilities, but it failed to deliver the expected efficiency:

Manual Updates: Any new service, update, or change required manual updates to the chatbot database, consuming significant time and resources.
Ineffective Query Handling: Only 30% of customer queries were resolved by the chatbot, with nearly 70% being escalated to human agents. Limited knowledge among human agents resulted in varying answers to complex customer queries, leading to customer dissatisfaction. Inefficiencies increased operational costs without improving the customer experience.
High Dependency on Human Agents: A team of 6 human agents was needed to handle unresolved and escalated queries.

Solution:

Our team was tasked with building a Proof of Concept (POC) to validate the effectiveness of an AI-powered chatbot. The solution utilized Generative AI (GenAI) and Natural Language Processing (NLP) to enhance the chatbot’s capabilities.

Phase 1: AI MVP (POC) Development (4 Weeks):

Integration: The chatbot was integrated with the client’s SMS API and a centralized knowledge base.
Data Collection: A two-week campaign encouraged customers to ask a variety of questions via SMS. These queries and their answers were aggregated to enrich the knowledge base.
Performance Validation: The POC demonstrated significant improvements, receiving positive feedback from the client’s call center and customers.
Phase 2: Full Implementation (3 months):

Automated Knowledge Updates: Enabled the chatbot to self-update with new information, eliminating the need for manual interventions.
Improved Query Handling: Enhanced chatbot performance allowed it to autonomously resolve 80% of customer queries with accurate, human-like responses.
Reduced Human Dependency: Reduced the number of human agents from 6 to just 1 for escalated cases, significantly cutting operational costs.

Results:

Quickfox technologies development team helped them to reduce dependency on human agents by 80%, cutting costs while maintaining quality. Automated content updates saved hours of manual work, keeping the chatbot current and efficient. The chatbot became capable of handling 80% of customer queries independently, with accurate and human-like responses. Positive feedback from customers and internal teams highlighted the improved experience. 

Overview:

The customer is a prominent commercial bank with over 150 branches, 100+ branchless banking units, and more than 200 ATMs.

The bank struggled with slow, manual retrieval of SOPs and regulatory updates, delaying responses and customer support. An AI-powered chatbot streamlined access, cutting response times from 10 minutes to instant, boosting efficiency and service quality.

Challenge:

Banks deal with a vast amount of information, including standard operating procedure (SOPs) manuals and frequent regulatory updates. These are traditionally stored in a Document Management System (DMS) and accessed via the intranet, requiring manual effort to locate specific information. This efficiency often leads to delays in resolving staff or customer queries, particularly when policies frequently change.

The bank faced difficulties in quickly retrieving regulatory updates, as staff had to manually sift through evolving policy documents.

Solution:

To address the challenges, we developed an AI-powered Internal Chatbot within 6 weeks. The MVP focused on streamlining document retrieval and providing instant answers to staff queries by leveraging cutting-edge AI and RAG-based systems. Here’s how the solution unfolded:

Phase 1: AI MVP development in 6 Weeks

Week 1: Requirement analysis and scope definition.
Week 2-3: Designed system architecture, ingested documents into ChromaDB for efficient knowledge retrieval.
Week 4-5: Developed and integrated the chatbot using Botpress for seamless interaction.
Week 6: Tested and deployed the MVP for live usage.

Phase 2: Future Enhancement 

While the MVP proved effective for documents in English and machine-readable formats, challenges arose with regulatory documents in local fonts. Current efforts are focused on developing custom OCR models to process non-UTF-8 encoded text and enhance the chatbot’s ability to handle diverse document types.

Results:

The AI chatbot transformed document access and query resolution:

Reduced Response Time: Previously, it took up to 10 minutes to manually search SOPs; now, responses are delivered almost instantly.
Improved Productivity: Staff now spend less time searching for information and more time on core tasks.
Enhanced Customer Support: Frontline employees are equipped with instant access to accurate information, leading to faster and more effective customer interactions.
The implementation of this AI-powered internal chatbot not only optimized operational efficiency but also established a foundation for ongoing improvements in knowledge management and staff productivity.