“
AI-Powered Chatbot and CRM
1
Introduction
Kiwi, an online travel agency known for its innovative approach to booking flights decided to innovate their take on customer support chatbot and enhance the client experience with AI.
2
The Need
The current chatbot struggled with complex prompts, leading to inaccuracies in responses. The primary client need was to reduce the burden on customer support by addressing as many customer requests as possible before resorting to direct support.
Simultaneously, focusing on streamlining the workflow, which was complicated by the missing integration with CRM, introduced additional challenges. Within the chatbot system, issues extended to the absence of contextual communication, leading to a uniform treatment of all users without considering reservations. Emotional recognition posed another hurdle, with the chatbot struggling to respond appropriately to customer frustration or stress.
The client sought a solution to enhance the chatbot's performance for a more efficient and personalized user experience.
3
The Challenge
The client's main challenge came from their previous solution. The POC delivered by Cleevio had a goal to transform their current solution into a fully AI-powered chatbot. We proposed that the chatbot responses should be evaluated using an accuracy score, with any response falling below a designated threshold triggering agent intervention.
Integration into other systems, such as CRM for order processing and invoicing, along with upselling capabilities for services like car rental and room booking, should be seamlessly implemented.
Additionally, the chatbot should aim to support multiple languages, recognising and generating responses in languages specified by users for a more inclusive communication experience.
*The client’s internal team overtook the development of the solution midway through the project.
4
The Cleevio Solution
Our innovative approach has resulted in a suggested potential AI-powered solution that empowers and optimizes customer support via the implementation of core features. Features below remained suggestions, as the client switched to in-house development before this stage.
AI-powered search for agents
The AI-powered search for agents ensures a faster and more relevant knowledge base search. AI is employed to retrieve documents and provide concise sourced answers, and understanding the existing knowledge base while chaining multiple smaller models refines queries and documents for increased efficiency.
Processing conversations in real-time
AI processes conversations in real-time for faster resolution and improved accuracy, analysing and providing relevant information. Automating repetitive tasks by AI reduces the workload for customer support teams, resulting in significant cost savings.
Summarize previous conversations
AI efficiently summarizes previous conversations, enabling a quick understanding of customer history for support agents. This streamlined workflow reduces information-gathering time, enhancing efficiency. Additionally, AI provides real-time assistance during conversations, offering relevant information and suggesting next steps.
Sentiment analysis & rephrasing feature
This feature offers real-time feedback on customer sentiment, enabling agents to adjust their tone for more effective support. AI's analysis provides insights into the customer's mood, helping agents better understand the need for tailored assistance. Predictive analytics leverage sentiment data to anticipate customer responses, enabling proactive issue resolution. Over time, AI's analysis of sentiment data identifies patterns for continuous improvement of customer service policies and procedures.
