CLIENT: VISA INDUSTRY: FINANCIAL SERVICES
Streamlining localized multi-channel marketing with an LLM constructor

Streamlining localized multi-channel marketing with an LLM constructor

“The Neurons Lab team demonstrated deep expertise in GenAI and worked diligently to build a tailored solution that has significantly improved our marketing personalization.”

Anna Melnychuk, Senior Marketing Manager VISA

 

Visa, the world’s leading global digital payments network, recently partnered with Neurons Lab. Together we empowered the marketing teams to craft localized, compelling messages for multiple channels faster and more efficiently.

 

Partner Overview

If you have a debit or credit card, there’s a good chance it’s one of the 4.3 billion cards issued by Visa. As the world’s leading digital payments network, Visa connects businesses, financial institutions, governments, and individuals across 200 countries, processing 283 billion payment transactions globally in 2023 alone.

Visa facilitates a wide variety of payment types – from P2P and B2C to B2B and G2C. Visa’s global network includes banks, governments, businesses of all sizes, and individual consumers worldwide. Therefore, its messaging must always be tailored to the target audience’s context and location to deliver maximum impact.

 

Project Overview

To help marketing teams overcome these challenges, Visa enlisted our help in creating a simplified constructor powered by Large Language Model (LLM) technology. The constructor would be available to marketing managers to streamline message creation, translation, and optimization.

To maximize marketers’ productivity, the generative AI module needed to be seamlessly integrated with its Campaign Atlas portal and allow for customizable parameter input (e.g., target language, channel, text length, and tone of voice selection).

The new tool had to support eight must-have languages and up to 11 nice-to-have languages and copy optimization for social media platforms (Facebook, Instagram, LinkedIn, Twitter), messengers (Viber, Telegram), email, and sales scripts.

Business Challenges

Before working with Neurons Lab, Visa had to tailor its messaging to each location and leverage various outreach channels. It is a very time-consuming endeavor as marketing specialists have to grapple with the following challenges:

  • Translating marketing copies accurately
  • Adapting them to specific communication channels
  • Ensuring they remain consistent in tone and messaging with the global corporate guidelines

This efficiency gap in marketing was the key challenge Visa turned to us with. The company wanted to leverage GenAI to increase the marketing teams’ productivity by tackling:

  • Language localization across 19 countries
  • Quality assurance to prevent spelling and grammar mistakes
  • Copy optimization for each marketing channel
  • Tone and messaging consistency, especially for smaller bank marketing specialists

 

Solution

The solution to these challenges was a simplified constructor that empowers marketing managers to streamline and optimize marketing messages using LLM technology.

The constructor includes pre-built templates and customizable parameters that users can leverage to generate effective messaging for various communication channels efficiently.

LLM architecture

The user inputs a query using the web app dashboard in the Campaign Atlas portal, including:

  • Query
  • Language selection
  • Tone selection
  • Channel selection

Next, the natural language processing (NLP) module built on LangChain receives the input and activates components for tone and channel augmentation, as well as translation.

Prompting options for tone include informative, authoritative, enthusiastic, inspirational, and so on. Channel choice includes social media, email, sales rep script, digital ad, and more.

Later in the chain, the NLP module handles translations. Initially, it leverages GPT-4, followed by custom models such as MarianMT and MBart from Hugging Face to improve accuracy and quality.

  • Near the end of the chain, the user receives several different intermediate output responses on the Campaign Atlas portal before choosing a text operation.
  • These post-processing text operations include requests to make the text shorter or longer, add emojis, simplify the language, improve the writing, etc.
  • This response enters the NLP module to perform the post-processing request, and the user receives the final output.

The platform stores a history of all intermediate and final outputs in an extensive database, allowing users to fetch past responses again in the future.

API and Cloud architecture

Neurons Lab implemented the LLM model using LangChain between the user inputs and the outputs within a LangServe wrapper—a Python framework well suited to APIs.

The main chain checks the input configuration, while the sub-chains handle tone, channel augmentation, and text customization. Suggestions are provided as the outputs.

To deploy the API solution, Neurons Lab used the AWS Elastic Container Service (ECS) alongside LangSmith for the prompt storage logs and evaluation.

This dashboard supports prompt management and the playground while providing exceptional log and chain tracking.

 

Results

The solution provides:

  • An integrated GenAI module: for NLP requests in the Campaign Atlas portal.
  • Customizable parameters implemented within the constructor: These included: translation to local languages, setting the tone of voice, and generating text for specific outreach channels.
  • In addition: Fixing spelling and grammar, adjusting text length, and simplifying the language.
  • Support for all eight must-have languages: Ukrainian, Kazakh, Moldovan (Romanian), Georgian, Belarusian, Arabic, French, and Russian.
  • Support for other languages including: Uzbek, Turkmen, Tajik, Serbian, Montenegrin, Macedonian, Kyrgyz, Albanian, Azerbaijani, Armenian, and Bosnian.
  • Optimization of text generation for outreach channels: Social media (Facebook, Instagram, LinkedIn, Twitter), Messengers (Viber, Telegram), email, and scripts for sales representatives.

 

About us: Neurons Lab

Neurons Lab is an AI consultancy that provides end-to-end services – from identifying high-impact AI applications to integrating and scaling the technology. We empower companies to capitalize on AI’s capabilities.

As an AWS Advanced Partner, our global team comprises data scientists, subject matter experts, and cloud specialists supported by an extensive talent pool of 500 experts. We solve the most complex AI challenges, mobilizing and delivering with outstanding speed to support urgent priorities and strategic long-term needs.

Ready to leverage AI for your business? Get in touch with the team here.

Get in touch Leverage the power of AI for your business.
Automating Business Operations with Generative AI in Cybersecurity
Peak Defence: Automating Business Operations with Generative AI in Cybersecurity
Elevating the world’s first open-source design platform with AI
Penpot: Elevating the world’s first open-source design platform with AI
Leveraging AI-powered recommendations for personalized engagement
Nection: Leveraging AI-powered recommendations for personalized engagement