Model Office

As part of their enterprise dashboard © Model Office wanted to create a chatbot solution to supply key information and guidance on all user questions. Providing a service which would be available 24 hours a day seven days a week. The chatbot would understand free text input so would need natural language understanding.

The © Model Office chatbot needed to be advanced enough to answer a variety of compliance and business development questions. These would be centred around her knowledge of a number of specific areas, for example information from the FCA code of business handbook.

Due to the constantly changing subject matter it also needed to be relatively easy to add or change the chatbots capabilities. The subject matter for this area has potential to be very large so the chatbot service needed to be scaleable.

© Model Office wanted to create a chatbot which could provide accurate responses to the large variety of regulary specific and general questions which would be asked as part of a complex enterprise level SAAS.

These questions could be simple FAQs such as:

User: "What are pension trustee duties?" MO: "Pension Trustees duties do not change across pension..."

Or more complex enquiries where MO would need to ask further qualifying questions to provide the most accurate information for the user enquiry e.g:

User: "What is the latest on Pension transfers " MO: "What concerns you? Pension misselling? Poor investor decision making?" User: "Pension Misselling" MO: "The FCA are doing some good work here Link to more information"

The chatbot would be provide as an API service which could be integrated into the © Model Office dashboard.

We were provided with some very useful intents and training phrases and user enquiry data from the team at Model Office as they know their subject matter inside out.

Intents were provided to us in spreadsheet format and from this information we were able to create the conversational scripts and then implement the conversation ability.

You can read more about how we run our chatbot projects here.

This was an iterative process. Matching user intents to core functionality and features and training the natural language processor to understand users and handle conversation failure scenarios gracefully.

The training data was invaluable for perfecting the bot conversations. The process highlights the need to monitor the chatbot as part of a continuous cycle of continuous learning.