AI Business Transformation Will Be Built On The Enterprise LLM
AI remains top of mind for many business and industry leaders, and the sector shows no signs of slowing down regarding investment and funding. Most recently, the generative AI startup Writer announced that it raised $100 million dollars and series B funding. What’s especially noteworthy about this announcement is that Writer positions itself as the “full stack-generative AI platform for businesses. Just this morning, Amazon announced that it has completely overhauled Alexa using generative AI LLM methods to drastically improve the ability of Alexa to understand conversational phrases context and respond appropriately. In the enterprise space, i.e., large businesses and organizations—AI advancements are setting the stage for a long and complex period of “AI business transformation” where technology stacks will be updated, new processes will be introduced, policies will be modernized, etc. And while new jobs will be created, many existing jobs will be redundant. In other words, like transformations in the past, such as e-commerce or social media—every company will become an AI company.
However, the AI business transformation may be even more significant than some of its predecessors. Let’s take the modern bank ATM, which automated many bank services and specifically eliminated the need to have as many human bank tellers. AI can potentially automate even more parts of the customer experience, specifically by using significant large language model technologies that could replace existing automated phone services and make it feel like you are conversing with a live human being—but you aren’t.
The Enterprise LLM
Enter the enterprise LLM or, simply put—a company’s version of ChatGPT. Companies are already beginning to build out their large language models. Unlike ChatGPT, which is built on deep learning models from what it can access off the Web, it is an organization’s proprietary data that is the foundation for building its enterprise LLM. What this means is that you could be interfacing with any one of a company’s customer touchpoints—phone, Website, chat, app and if a company has built out its version of a large language model with its data—you can be conversing with that large language model which may already know who you are, your history with the company, what products or services you have already bought and what you may want to buy next. A company that seeks to build its large language model will no doubt have to invest in new infrastructure technology and reprioritize initiatives. Still, an enterprise LLM consists of a secure cloud service, a deep learning model, data sources, and the experience layer in which a person interacts with the LLM.
The Hyper-personalized Customer Experience: These kinds of hyper-personalized experiences are already becoming part of the consumer experience made possible by generative AI. Auto retailer Carvana recently experimented with combining key customer data points such as the buyer’s name, the car they bought, and where and when they bought it. The initiative is playfully called “Joyride”. Using this data, they generated thousands of pieces of content that told individual stories around this data:
“We know that our customer, Holly, bought a Carvana Corvette on September 29, 2019. September 29 is National Coffee Day, and everyone was talking about the ending of Game of Thrones (it… certainly was an ending). Our system combines that knowledge with AI art, animated assets, and stylistic parameters to weave those elements into our AI-generated video.”
Carvana now sits on “five years” of personalized video content that can be shared with the same customers who inspired that content with the data they contributed to as part of the car buying experience.
There’s an LLM For That
Some of the world’s largest consultancies have already built out proprietary enterprise large language models and are expanding and refining them. McKinsey already has employees testing out its proprietary LLM named “Lilli,” and it is slated to become more widely available this fall: “The interface will look familiar to those who have used other public-facing text-to-text based gen AI tools such as OpenAI’s ChatGPT and Anthropic’s Claude 2. Lilli contains a text entry box for the user to enter in questions, searches and prompts at the bottom of its primary window, and generates its responses above in a chronological chat, showing the user’s prompts and Lilli’s responses following.”
What a proprietary LLM can do, which “consumer” LLM products such as ChatGPT, Bard, etc., cannot, is access specific data that isn’t publically available to the AI Web crawlers used by sources such as Open AI. Companies are already taking advantage of putting code in place on their Websites to block these crawlers, which are only for the information in the public domain. Many companies already possess valuable customer data, which is already tapped into as part of the customer experience but not through an LLM. However, any existing interface that serves up customer data has the potential to become part of an organization’s custom-built large language model, and just like we have an app to do almost everything—a company can have an LLM interfacing with customers as AI assistants, guides, customer service agents, etc.
The Conversational Interface Is Coming
Back to the example of early ATMs… We’re already accustomed to what the UX world calls “GUI” or Graphic User Interfaces. We use them on our phones, at kiosks, etc., etc. They are the primary language we know and understand regarding technology interfaces. LLMs will create another primary interface that will become just as prominent, the “CUI” or the Conversational User Interface. Presto is one of the first companies working with several fast food chains leveraging AI technology to automate the drive-through experience. What a customer will do in these instances is talk to an “AI worker” via voice while following their orders being updated on a screen. This is an excellent example of GUI and CUI working together. However, it’s still early, and future conversational experiences will feel more natural and less robotic as AI technology progresses. Other players, such as Air AI, are working on this more natural, conversational version of customer support; directionally, these solutions add fuel to the fire in the debate on automation, jobs, and the economic impact of these things.
Conversational Experiences and the Inevitability of the LLM Engine
Inevitably, nearly all companies will eventually be running their own proprietary large language model, either complemented or co-built with one of the large public LLMs or existing independent of. Proprietary enterprise LLMs will be the foundation for companies to expand their customer and brand experiences into the conversational and hyper-personalized realm. Interacting with a company will feel like having a friendly conversation with someone who knows a lot about you based on the company’s data—or if your eyes simply need a break from one of the many screens we interface with all day long, a company’s LLM may just turn your customer experience into a friendly chat.
Read the full article here