Opinions expressed by Entrepreneur contributors are their own.
GPT-4 and similar technologies are making possible ideas businesses only dreamed of just a few months ago. Such dislocations in technology allow early adopters to massively outcompete those that come late. These new technologies are poised to impact the competitive landscape significantly in the near term.
Like many, I first started using ChatGPT and OpenAI’s APIs in the last several months. What at first seemed like a very promising new technology in search of a use case has already completely transformed my software business and how we think about our future. In this article, I’ll discuss some of the practical use cases for this new technology to supercharge your business.
You don’t have to be particularly sophisticated to see the potential of Large Language Models (LLMs), the type of model that powers ChatGPT. If you watch the news or open any social media application, you’re flooded with input prompts and gimmicky use cases for this new technology. Most of the posts I’ve seen focus on having ChatGPT create text based on concepts it knows, mainly from web data. Ask it to write a poem, a short story, a legal document or produce code, and it does so with incredible speed and quality. Where things get even more interesting is when you start using it to analyze data that it wouldn’t normally be able to access. Because everyone has access to LLMs, competitive advantage will come from a company’s ability to pair it with novel data that isn’t broadly available.
Let’s discuss two use cases that have been particularly powerful for us at Nomad Data, a software platform that allows companies to describe the data they need for a project and be connected to vendors that have it.
Related: GPT-4 Is Just the Tip Of the Imminent AI Revolution: Sam Altman
Sales prospecting
LLMs will completely transform sales targeting. As an enterprise software company, we use many channels to find new clients. Social media marketing and email marketing have been two particularly useful methods for us. The way to be successful with both is pretty straightforward: Send the right message to the right group of people, preferably at the right time. From our experience, the hardest part to do well is finding that right group of people.
If you use email, for example, most list-building services allow you to filter business contacts by company size, title, industry and other high-level metrics. For some industries, this is enough. At my company, our issue is that we target broad roles such as Data Scientist or Data Engineer. Ninety-five percent of data scientists are not the right target for us, but since filtering is done at the title level, this is the best we can do. The same lists of targets are then used for social media targeting, leaving us with the same issue.
LLMs present a completely new opportunity for targeting. If I were to look at a data scientist’s social media profile, I can quickly tell if they are actually a good target for our software. If I had the time and patience to do this with every data scientist’s profile, I could come up with an amazing target list, massively accelerating the speed at which I can get my message to the right people. In turn, I’d be dramatically reducing my marketing costs since I wouldn’t market to anyone outside of my focus area.
LLMs allow you to do just this — and at scale. You can feed in a profile and teach the LLM what about that profile makes it a good or bad fit. You can do this thousands or millions of times to produce the perfect lead list. The next question is always, “Well, where can I get this data?” This is a common dataset that you can find online or through our data discovery service at Nomad Data and purchase.
Related: What Does ChatGPT Mean for the Future of Business?
Data entry
Data entry is a common problem for most businesses. You have a large volume of data you need transformed to make it more usable. This might be removing pricing data from a large set of contracts, or it might be removing key pieces of information from a long product description. Common solutions to this previously were to hire armies of people to go through the data one piece at a time and enter the new information into another system. Another novel solution was to use a service like Amazon Mechanical Turk to essentially outsource these small tasks to thousands of people, paying a small amount per completed task. With LLMs, you can now automate this process, transforming data at a pace that just wasn’t possible with human intervention.
At my company, we’re using this approach to perform many different tasks. One such task is to improve the profiles of vendors participating in our marketplace. Because we have thousands of partners, it wasn’t feasible to improve each vendor’s profile by hand and ensure the quality was up to our standards. With LLMs, we can train the model to understand what makes a profile good or bad, and then in seconds, it can let us know which ones aren’t up to par. It can even produce useful feedback which we can share with the vendors to help them improve their profiles.
We’re also in the process of adding similar functionality within our application that allows users to input free-form text. We can use this approach to ensure the quality of what is being inputted is sufficient without any manual intervention.
Related: How to Use AI Tools Like ChatGPT in Your Business
Create a competitive advantage before you fall behind
These are a few of the simpler use cases we’ve come up with for LLMs that all rely primarily on data that your company possesses or can acquire. We have many more we are researching that are even more transformative.
The most interesting use cases involve inputting unique data with very detailed sets of instructions on what task the AI should perform. These instructions get built into the prompts, which with GPT-4 can be quite long, allowing for a significant competitive advantage to be derived from using them well. On our data discovery platform, we have already seen an increase in demand driven by clients looking for data to power these types of models.
Throughout my career, I’ve never seen a technology have such a profound impact in such a short period of time. Inventions like the semiconductor, cell phones and the internet took well over a decade to start having a material impact. This pace allowed businesses to prepare for the coming change. LLMs are going to completely change the art of the possible, and not in decades, but in months. Finding ways to create an advantage by using them is going to be essential to maintain competitiveness.
