Case Study: How do I improve colleague access to relevant knowledge?


Following data insights analysis by Heron for one of our banking clients, a major take away was low customer satisfaction levels around regular payments such as; standing orders, direct debits and ad-hoc payments.  This led the bank to investigate how branch and call centre colleagues access information to help them answer client questions.

Aside from asking more experienced colleagues, who are typically unavailable during busy periods, colleagues can access the Bank’s internal knowledge base with ‘google-esq’ search requests for information.  This action returns a sizeable response that is difficult to consume in a timely fashion when serving a customer.  This led colleagues not to use the knowledge base at all, instead resorting to ad-hoc measures they created themselves, with mixed results and a questionable customer experience.

The Bank asked Heron to recommend a solution that would allow colleagues to ask questions in a conversational manner to provide the colleague with answers for the customer. The solution was trained to ask supplementary and clarification questions in order to provide best practice answers and discrete processes for the colleague to follow.

How did Heron approach this?

Heron proposed a chatbot solution using IBM’s Watson Conversation and Rainbird to understand the questions being asked to provide the necessary interaction and answers for the questions asked. Watson provided Natural Language Processing capability, allowing the system to understand intents of the question, while Rainbird provided the answers.

Using a mixture of the Bank’s collateral and access to SMEs, Heron worked to understand how clients and colleagues ask questions so as to train Watson Conversation and build the Rainbird knowledge map.

Technologies Used:

  • Machine Learning NLP/NLU
  • Chat interface
  • Automated decision making


A front-office chatbot for colleagues providing rapid, contextualised recommendations and answers to questions with over 90% accuracy. This led to improved colleague and customer satisfaction.

The bank is now considering additional topics to expand the value of the chatbot.