The AI Gold Rush
The AI gold rush is underway, as highlighted by mentions of the tech in earnings calls and Nvidia’s stock chart. Naturally, businesses want to take advantage of a new opportunity, but it’s a complex path. Liberty Advisor Group recently published an article highlighting how lessons learned from the actual gold rush of 1849 may apply today, and this paper seeks to expand upon the concept by outlining tangible AI applications, including their underpinning technologies.
Predictive vs Generative AI
At a macro level, AI technologies can be considered in two general use cases for enterprises: predictive and generative. There are dozens of rabbit holes within both categories, from data format to model selection. However, considering just those two buckets at the C-suite level facilitates strategic brainstorming.
Predictive analytics use historical data to predict future outcomes. These models can inform a business of how much material to order, when machines will break down, or which customers will likely grow in revenue.
Predictive Machine Learning Case Study: Carvana
Carvana was an early adopter of machine learning technology – they hosted a significant Kaggle modeling competition in 2011 [1]. Kaggle is a platform where data scientists compete to create the best-performing predictive models after being provided a training dataset. In this scenario, the winning team got a cash prize, and Carvana got a linchpin algorithm to boost their business.
In an excellent example of the scale required to complete this type of work, Carvana provided a dataset with 32 independent data points for 48,707 cars. Data included year, make, model, purchase location, odometer reading, etc. The dataset also provided a binary field called ‘BadBuy’ – if a car was a ‘bad buy’ in real life, this was annotated. Therefore, the vehicle in that row of data had ended up being a poor purchase by Carvana, considering its eventual resale value.
With this information, data scientists used a small chunk of that data to train a model; this process is where a machine learning mechanism such as a decision tree or neural network (with numerous variations) learns what factors most heavily contribute to a car becoming a ‘bad buy.’ Next, with the remaining, larger chunk of data, the modelers remove the ‘BadBuy’ data field and ask their models to predict whether each car, row by row, would become a ‘bad buy.’ They then compare those predictions to what actually happened to figure out how accurately the model performed, making countless tweaks to improve the predictions over time.
Carvana used this data-driven, machine-learning method to improve their decision-making process when buying cars at auction, seeking to avoid unprofitable vehicles.
Generative AI
Generative AI creates new material in various forms by drawing from massive training datasets. Enterprises have used generative AI to enhance customer support channels, personalize marketing efforts, and provide custom, labor-saving assistance to their employees.
Generative AI / LLM Case Study: Roadside Assistance
A recent client in the roadside assistance industry asked Liberty Advisor Group to help implement an enhancement to their call centers, which handle the critical task of dispatching a roadside service vehicle to stranded clients, such as a tow truck. As you can imagine, if you’ve ever sat stationary on the side of a highway, the customer’s wait time is a critical measure of success.
The business sought to free up call center agent time by routing some customer traffic through an online chatbot utilizing a Large Language Model (LLM), a form of generative AI. The model’s training on millions of words and sentences gave it the foundation to produce grammatically correct, rational answers to word prompts. Then, the business further enhanced it by inserting key data points from their operations: what services a customer could dispatch themselves from the chatbot, situations to immediately escalate to a live agent, and integrations to a backend platform that allowed customers to retrieve information about their insurance policy.
This method of customizing an LLM to fit the needs of an individual business problem takes advantage of the baseline AI technology while putting reigns around the chatbot to give the customer a guided path.
Where AI Fits into Your Business
To figure out how to apply these hyper-efficient technologies to a business, Liberty Advisor Group recommends that you ask the following questions:
- What type of business problem is being solved? Are we predicting something or generating something?
- For predictions, do we have historical data to train a machine-learning model? Are we storing new data week after week?
- For generative models, what additional information would the model need about our business to be useful? Are there IT systems that make sense to integrate?
- Lastly, do we have the data infrastructure and personnel on staff capable of developing this functionality?
These foundational prompts may lead to dozens of follow-up questions and tasks before execution. Still, at the strategic level, it’s critical to find the proper fit within your business for a technology-based enhancement to provide value.
How Liberty Can Help
Liberty Advisor Group understands that data and processes are the lifeblood of AI systems. Furthermore, we recognize the importance of having a well-defined strategic AI roadmap. Our team works closely with your organization to develop a clear and actionable plan for AI adoption. We align AI initiatives with your business goals and guide you through the entire process, from inception to deployment.
Learn More About Data Services ➞
About Liberty Advisor Group
Liberty Advisor Group is a goal-oriented, client-focused, and results-driven consulting firm. We are a lean, handpicked team of strategists, technologists, and entrepreneurs – battle-tested experts with a steadfast, start-up attitude. We collaborate, integrate, and ideate in real-time with our clients to deliver situation-specific solutions that work. Liberty Advisor Group has the experience to realize our clients’ highest ambitions. Learn more on LinkedIn and Twitter.
References
[1] “Don’t Get Kicked!,” Kaggle, https://www.kaggle.com/c/DontGetKicked
