Next Best Offer

Enable banking sales professionals to generate customized yet automated follow-up messages with the power of Dataiku, classic predictive machine learning, and Generative AI.

Banking professionals’ time is valuable, and clients now expect to be engaged only on relevant offers and with relevant messages. This large language model (LLM)-enhanced “Next Best Offer” application allows employees to quickly generate high-impact emails and integrate these in their customer engagement strategies. 

Feature Highlights

  • Fast & Customizable: Instantly generate a customer email, then further refine that email with instructions in natural language (e.g., “make the email less formal” or “refer to our conversation from last Thursday”) — even by voice.
  • Personalized: Quickly add personalized recommendations for different offers from the bank in the email. These recommendations come from a predictive model (also built in Dataiku) and, importantly, are not AI-generated.
  • Composable: Powered by customer data, existing predictive models built and running within Dataiku, and Generative AI. 
  • Multi-lingual: The model provides the same high level of performance across many languages (one of the advantages provided by LLMs) — perfect for banking professionals that speak or manage customers in more than one language.
  • Controlled: Banking professionals have final oversight of the message content before it is sent to the customer, including the ability to add a (non LLM-generated) disclaimer that has been approved by the legal department directly in the application.
  • Secure: Banking organizations can select their choice of containerized LLM or private LLM to quickly deliver this use case while preserving full privacy of their customer data and next best offer recommendations.

How It Works: Architecture

Dataiku’s LLM-enhanced “next best offer” app connects to enterprise data environments, including customer data. Dataiku gives you full flexibility to configure the application the way you want, using the data and systems that you want to connect to.

In this example, the app is taking its personalized recommendations from an existing “next best offer” model. This is a traditional machine learning model, also built in Dataiku, that makes recommendations for what to offer customers. The bank doesn’t trust the LLM to come up with an accurate prediction like this, so they provide the LLM with the recommendation they would like it to include in the email. 

Responsibility Considerations

The LLM-enhanced “next best offer” app from Dataiku allows the banking professional to add a legal disclaimer to the end of the email. This is a pre-scripted disclaimer that has been vetted by the bank’s legal department; it is not being generated by the LLM.

In addition, the app is designed to be human-in-the loop, meaning the banking professional has final oversight of the message content before it is sent to the customer. 

Other recommendations for the responsible use of a LLM to generate personalized messaging fueled by “next best offer”:

  • The next best offer system should be audited to ensure it is making predictions free of bias — Dataiku makes this simple with explainable AI features.
  • The organization should have an overarching Responsible AI policy, and it should be enforced with both the next best offer system as well as its LLM-personalized content generation extension.