Co-founder Sheth on Gupshup's ACE LLM suite that enhances conversational messaging

In today's business landscape, automated conversations play a crucial role. Yet, aligning artificial intelligence’s (AI) potential with practical business requirements poses its challenges. Gupshup, a conversational messaging platform, recently introduced ACE LLM, a suite of GPT models fine-tuned for business discussions.

In an interaction with Beerud Sheth, co-founder of Gupshup, we delve into ACE LLM's inception, its integration, benefits for enterprises, the role of data and the ripple effects it creates. Edited Excerpts:  
 
What was the motivation behind developing ACE LLM?   
   
Our motivation was to meet customer needs and our own vision of enhancing customer support. We aim to facilitate enterprises in customer engagement through channels like WhatsApp and chats. Effective automated conversations play a crucial role, needing intelligence, seamlessness and engagement. The rise of AI and LLMs like GPT-3 and GPT-4 has improved conversation quality. However, these models can’t be directly used, especially for business talks requiring accuracy, context, relevance, and goal orientation.

For instance, asking GPT-3 or 4 about Citibank credit card interest might yield inaccurate results, a problem in business conversations. Similarly, if an insurance company receives a pizza-making query, irrelevant answers aren’t suitable. Enterprises want to use LLMs, yet they fall short. We designed ACE LLM to bridge this gap, offering what foundation models lack to create exceptional customer experiences. This approach fills the void between foundation models and enterprise requirements.   
    
How does ACE LLM work with your company’s conversational engagement platform?   
   
Think of it as the powerhouse of a car or an airplane. Our conversational platform processes user queries, using various tools to formulate responses, maintain dialogue, gather user data, and personalise interactions for marketing, upselling, customer support, and more. This is where our engagement platform shines. Now, integrating the ACE LLM takes this to another level. Conversations become natural, handling more diverse queries, and greatly enhancing the customer experience. It's like upgrading a car from a small engine to a turbojet — suddenly, it’s faster, more agile, and capable of so much more. The ACE LLM significantly enhances our platform’s capabilities.
 
How enterprises can benefit from ACE LLM, can you give us some use cases?   

Enterprises can reap significant benefits from ACE LLM. To understand why, let’s focus on the core objectives of enterprises. Essentially, enterprises aim to foster customer engagement. This involves cultivating strong customer relationships, addressing inquiries, and providing a seamless experience. To illustrate, envision a scenario: think of a shopkeeper and a customer entering a store. Even before the customer arrives, the enterprise can proactively notify them about new inventory or ongoing festival sales. Once the customer is present, personalised assistance can be offered — understanding their preferences, suggesting suitable items, and facilitating transactions. This encompasses the entire customer journey, spanning marketing, commerce, and support.

Here’s where ACE LLM steps in. Unlike traditional chatbots that follow a rigid script, ACE LLM offers a dynamic and fluid interaction. It empowers customers to converse naturally, like seeking recommendations for a gift within a specific price range. This AI-driven approach elevates the customer experience to new heights. It enables effortless browsing, searching, and querying, all seamlessly powered by AI. As a result, enterprises can enhance their engagement strategies across every phase of the customer lifecycle, ensuring a more delightful and attractive interaction.
 
Did data scarcity pose a challenge for you when building ACE LLM? 

Clearly, the foundation of AI is data. To embark on any AI endeavor, the initial question to address is, “What data do we possess?” Gupshup, having a 15-year history in the field, has collaborated with diverse sectors such as banking, retail, education, healthcare, and more. This extensive experience has generated substantial historical data within these domains. With a track record of sending out billions of messages annually for a decade or more, Gupshup has amassed a vast repository of information — amounting to 100 billion messages per year. This rich dataset underpins the development of our models. Importantly, we are leveraging our substantial expertise and existing data; we are not venturing into unfamiliar territory.
   
What kind of impact are you looking with ACE LLM?   
   
We believe this will have a significant and positive impact. The reason is simple: by delivering an enhanced customer experience, superior technology, and advanced capabilities, we expect a substantial boost. Existing customers will seek more features, driving increased revenue. Moreover, new businesses will become customers to access these benefits, expanding our user base. This growth in customer numbers and higher spending per customer will naturally elevate our overall company revenue.
 
Another noteworthy point is Gupshup's consistent innovation in this field. Our tech-centric approach and strong engineering teams consistently position us as pioneers. This not only reinforces our leadership but also contributes to revenue growth. Importantly, real customer usage yields invaluable feedback for enhancements.
   
What enhancements or additions we can expect from ACE LLM?   
 
AI usage involves various aspects: accuracy, core knowledge, security, compliance, and data residency (keeping data within specific countries). We’re committed to enhancing speed, quality, and affordability. This innovation isn’t limited to just ACE LLM; we’re part of an evolving ecosystem with improving hardware, emerging algorithms, and new foundation models. Indian engineering is ascending to new heights, foreshadowing further developments in domain-specific AI models.