Technologies:
About the Client
Customer is a successful manufacturer of grocery and food products that supplies its products all over the world. One of its department runs marketing campaigns on the regular basis.
Business Context
To run the marketing campaign, customer has to create a large variety of different internal and external marketing assets, such as talking points, emails, blog articles, social media posts for various platforms. Those assets has to match specific styleguides and include key campaign messages. Moreover, most of the assets has to be translated into different languages for local marketing departments.
Amazinum’s team of Data Scientists had to automate marketing assets creation and translation.
The work of specialists was mostly related to LLM systems in terms of chatbots and the internal environment of employees.
Among the tasks that were set before our specialists were:
- Develop templates for corporate mailings that would be generated by LLM
- Support for user conversations with LLM
- Replacing specific acronyms with their full form using custom company glossary to simplify the assets for newcomers
- Develop an LLM model that would generate posts for social media
- Develop an LLM model that would analyze and repeat the writing style of the company’s employees
- Develop a service for translating documents in different file formats into multiple languages, keeping the original formating and style
Amazinum Data Scientists in Action
To fulfill our client’s first task, specialists chose the Google LLM model PaLM 2. Using the LangChain framework, data scientists contacted and connected to the model and fine-tuned it. After that, the Amazinum team moved to the industrial design stage, as it was necessary to receive a specific output from the model. For this, we used Google services – Cloud Build, Cloud Functions, and Vertex AI.
Amazinum data scientists used the LangChain framework to support end-user conversations with the chatbot. Because it has integration with the Vertex AI service, it is convenient to use for chatbots and works with programs that use large language models.
Deciphering acronyms required writing specific prompt design requests. For this, our specialists used few-shot technique alongside with in-context learning.
The generation of posts for social networks such as Instagram, Facebook, and LI required the formation of the same prompts, which were run through the LangChain framework and the input document of the post template.
To generate posts that would duplicate the existing style of expression in newsletters, replies, or social networks of the company, Amazinum specialists prepared large amounts of data provided by the client’s company. Model were fine-tuned with data on the developed Vertex AI pretraining pipeline. In this way, we received a ready-made model that could work with end users, fully duplicating the company’s style.
Outcome
A product-based company can benefit greatly from the deployment of a domain chatbot in a number of ways, including increased customer engagement, operational effectiveness, and overall business performance. Here are some benefits that our client received:
Enhanced Customer Assistance:
Customers can get immediate help with product inquiries, order tracking, and troubleshooting through a domain chatbot. Improved customer service, quicker response times, and higher levels of satisfaction are the results of this.
FAQs and Finding Information:
Chatbots are excellent at responding to commonly asked queries and offering information on usage, specifications, and product details. This guarantees that customers can access information whenever it’s convenient for them and lessens the workload for customer support teams.
Enhanced Sales Conversion Rate:
Chatbots can assist customers with a smooth purchasing experience by guiding them through the process and answering any questions they may have in real-time. Higher sales conversion rates may result from this.
Event and Promotion Notifications:
Companies can use chatbots to tell clients about forthcoming special events, promotions, or new product launches. Proactive communication has the potential to generate customer interest and increase involvement in promotional activities.
Customization
Businesses can help businesses better target their marketing messages and product recommendations by using chatbots to have interactive conversations with users and learn about their preferences.
Brand Loyalty
Using chatbots to provide customers with a customized and attentive experience can boost brand loyalty. Customers who are happy and involved are more likely to make additional purchases from you.
Cost-Efficiency and Scalability
Chatbots automate repetitive customer interactions, which lessens the effort for human support teams and saves money. Furthermore, chatbots are scalable, meaning they can manage an increase in customer inquiries without requiring a corresponding rise in resources.
