About
Revolutionizing Home Insurance Sales with Conversational Intelligence. Our AI-driven platform analyzes real-time conversations, providing agents with tailored insights for personalized offers, increasing sales and enhancing customer satisfaction.
Innovation presentation
Orange has entered the insurance market in Spain by leveraging their expertise in data analytics, digital capabilities, and customer relationships. This has allowed us to offer innovative insurance products and services that cater to the evolving needs of consumers.
To make this possible, we have deployed a platform that is providing us with high integration speed and scalability with insurance companies, allowing us to adapt to new ways of buying and selling insurance in the world of the new digital society. The platform integrates Orange's distribution channels and insurance carriers' products.
Furthermore, we have been one of the first mediation agency in Spain that are using Generative AI to enhance the Home Insurance Sales Experience and Efficiency.
Improving Home Insurance Sales Experience:
In a commercial context, conversational intelligence encompasses all the tools, techniques and technologies for analyzing and processing the interactions between the agent and the potential customer. In case of a sale, it is defined as an agent ability to evaluate and analyze conversations and exchanges with the customer to extract key insights, adapt strategy, and ensure that we give the customer what they really need.
Our project, thanks to Generative Artificial Intelligence, allows us to analyze the conversations held by the agent in the home insurance’s sale, transcribing and analyzing each one of them to make improvement recommendations and identify opportunities.
In addition, this analyzed and extracted information is fed directly into our platform, providing the most accurate information to our client to better understand them and give them what they really need. For example, in a conversation with a client or potential client, it could be glimpsed that they have tenant's home insurance that expires in 8 months and that they have it with company X, with some coverages that interest them and others that they do not. What the project allows is to extract all that "conversational intelligence" from the call and automatically, prior request for consent, inject the most interesting data into the platform, (the sales agent does not have to do anything).
This way, the agent will be provided directly with the best and most adapted information for the sale, and the customers, prior to the expiration of their home insurance, will be made the best offer adapted to their needs.
To carry out our project, we first did a pilot to analyze 30,000 transcriptions from a Call Center, 20,000 customer conversations, 7,000 Outbound Calls and 13,000 Inbound Calls.
An infrastructure has been set up. We have collected all the records, transcribed them to text with WhisperUI, as OpenAI Whisper is known for its high accuracy, and analyzed the information with an LLM – Claude 2 to check the following requirements to improve the experience of the customer:
- Reading the legal terms to the customer and informing them that the conversation is going to be recorded
- Competitor identification, pricing and expiration, so we can better understand what they have to offer our customers the best suitable option for them
- Patterns between calls that end in sale and non-sale
- Product & coverage explanation through examples
- Number of calls required to finalize a sale to reduce them, as well as the best moment of the day to call them
Improving Home Insurance Sales Efficiency:
On the other hand, one of the main keys to managing a sales team properly is to have the most important statistics and data on the activity. A key part of keeping track and ensuring a proper business pitch is reviewing calls and recordings. Thanks to AI, we can, in many cases, reduce this work of several hours to a task of a few minutes and in other cases do things that were previously incomprehensible.
Thanks to the pilot mentioned above, we obtained relevant data regarding the sale of home insurance.
Evaluated metrics and working method:
1. Metric selection:
• Traditional metrics surrounding agent efficiency
• Metrics based on generative AI:
o Conversation structure
o Expiration and current policy company
o Legal text comprehension (conversation quality)
o Coverages
2. For each topic, insights are determined and evaluated to generate a quantitative assessment of the results.
3. For each topic, an initial prompt is created.
4. The model that best responds is selected. In this case, the model selection (Claude 2 – LLM) is based on the length of the conversations to be analyzed. With Claude 2 we have been able to input up to 100K tokens in each prompt, which means that Claude can work over hundreds of pages (there are very long calls).
5. An iterative process is carried out to improve the prompt
6. Finally, responses are obtained for the test conversations, and a quantitative assessment of them is conducted to train and validate the model.
After the pilot, it has been operationalized by integrating the relevant information into the sales platform to:
- Ensure compliance with the law – Customer protection
- Ensure that our agents are meeting our quality standards
- Adapt to our customer needs by finding the best moment of the day to call them
- Simplify the understanding of the coverages through examples
- Offering the best solution tailored to our customer needs
Uniqueness of the project
We have been pioneers within Orange in overcoming existing barriers by meeting all the requirements within a large corporation, and the results of which have been a source of inspiration for other areas of the company.
In addition, within the framework of the pilot and its extrapolation we have achieved a great time to market to be able to put the project into production, making us one of the first mediation agency in Spain using Generative AI to enhance the Home Insurance Sales Experience and Efficiency.
