Cas client
Alternative banking service for customers
Deploying a Conversational Agent to meet customer needs across corporate and support platforms.
À propos du client.
A French alternative to traditional banking — no income requirements, no overdraft or credit, and already trusted by over 3 million customers nationwide.
What was your approach to the issue?
Talisker’s involvement in the project was positioned at the intersection between business teams and Amelia’s IT teams, who were responsible for implementing the conversational AI solution. The mission was structured around three key workstreams.
The first workstream focused on identifying use cases and designing the AI’s conversations.
This involved drafting functional specifications for the AI dialogues in close collaboration with client representatives, for handover to Amelia’s IT teams. Talisker also ensured coordination with the support teams managing the help portal, to validate the relevance of responses and, where necessary, suggest adjustments to business processes.
The second workstream dealt with IT coordination and version management.
This included overseeing Amelia’s IT teams to implement updates, conducting testing, following up on feedback to ensure the AI’s ongoing improvement, and finalizing technical documentation for the project.
The third workstream focused on language understanding and training the AI.
This required a strong grasp of business needs in order to design the intent architecture (categorizing types of customer requests) and build the validation dataset (VDS), which was then used to train the AI effectively.
what was the key to your success?
The client’s strong commitment to adding a new contact channel was the key driver of our success. The project manager’s involvement in the conversational AI aspect was also crucial in ensuring the successful completion of the mission.
Our shared experience with Amelia further streamlined communication and enabled us to bring in the necessary expertise at the right moments to accelerate the project’s development.
Collaboration with various stakeholders from the client side allowed us to centralize all the essential elements for a smooth implementation, ensuring we accurately met the business needs.
Where did you start from?
The project was initiated in the context of increased customer service demand. To address this challenge and adapt to a new type of clientele, the client aimed to diversify its contact channels.
What were the results?
The implementation of this conversational AI was carried out in several stages, resulting in the delivery of different versions for the client’s institutional website.
Version 1 enabled the AI to intelligently redirect users to relevant FAQ articles and open a form based on the user’s request. The AI was deployed on both the support site and the main site, under the “Contact” section, and was also accessible via Facebook on weekends.
Version 2 expanded the AI’s capabilities by specializing the form creation process, allowing it to accurately identify user requests and adapt the associated business workflows.
To carry out these tasks, a team of four people was assembled, including two business resources and two AI specialists.
