What Is Generative AI, and How Can Tools Like ChatGPT Help Businesses?

Beyond the headlines and the hype, though, what does the rapid emergence and interest in tools like ChatGPT mean for businesses? Experts say that these generative AI tools may be able to augment human capabilities in a wide range of industries.

The tools can often accelerate learning for employees by tapping into large knowledge bases and providing quick answers. And while experts don’t think ChatGPT and tools like it will do away with millions of jobs, experts do think the technology can competently handle basic tasks in many industries. Not only will that save users time, but it also will provide a framework within which humans can work creatively.

What Are Generative AI Tools like ChatGPT?

Before looking ahead, it’s worth unpacking what this technology is. ChatGPT is an application of generative AI, which is “a branch of computer science that helps people create new content from previously created content, such as text, audio, video, images and code,” says Ritu Jyoti, group vice president of worldwide AI and automation research at IDC. This is usually in response to short prompts.

ChatGPT is the most high-profile of these applications. DALL-E, another OpenAI product that received attention last year, is a similar tool that generates images in response to prompts.

DISCOVER: Find out how AI is changing the customer experience.

How Does Generative AI Work?

Generative AI is built on machine learning models, the first of which were “trained by humans to classify various inputs according to labels set by researchers,” as a McKinsey article notes. For example, a model could be trained to label social media posts as positive or negative; in this example, the human is teaching the AI how to label.

“The next generation of text-based machine learning models rely on what’s known as self-supervised learning,” the McKinsey article explains. “This type of training involves feeding a model a massive amount of text so it becomes able to generate predictions. For example, some models can predict, based on a few words, how a sentence will end. With the right amount of sample text — say, a broad swath of the internet — these text models become quite accurate. We’re seeing just how accurate with the success of tools like ChatGPT.”

ChatGPT is what is known as a large language model, with billions of parameters, trained on a large data set and then fine-tuned to produce specific outputs in response to text prompts.

What Is GPT-3.5, and How Does It Power ChatGPT?

GPT stands for generative pretrained transformer, a technology developed by Google researchers. A transformer is a neural network that learns context and meaning in sequential data by tracking the relationships between words.

These models are trained on a broad set of data, including academic articles, federal and state legal opinions, patents, websites, social media posts, general facts from sources like Wikipedia, books, open-source code and other resources. They then use that data to predict what the next word in a sentence is most likely to be and use that calculation when providing responses.

ChatGPT reflects OpenAI’s attempts to continually improve the GPT model. Originally based on GPT-3.5, ChatGPT Plus (a premium version of the tool) was recently upgraded to support GPT-4 and uses a significantly larger knowledge base that is capable of interpreting images.

“When shown a photo of a boxing glove hanging over a wooden seesaw with a ball on one side, for instance, a person can ask what will happen if the glove drops, and GPT-4 will respond that it would hit the seesaw and cause the ball to fly up,” a Washington Post piece noted.

GAIN INSIGHT: Find out how to prep your business for AI applications.

What Are Some Notable Use Cases for GPT-3.5 and GPT-4?

There are a wide range of use cases for GPT-3.5 and similar models. Such tools “stand to unleash more creativity into the world by prompting humans with starter ideas,” write McKinsey partners Michael Chui, Roger Roberts and Lareina Yee.

The authors point to a broad set of applications across functions, including but not limited to: 

• Marketing and sales, for crafting personalized advertising, social media and technical sales content (including text, images and video) and creating assistants aligned to specific businesses such as retail
• Operations, to generate task lists for efficient execution of a given activity
• IT/engineering, for writing, documenting and reviewing code
• Risk and legal, to answer complex questions, pull from vast amounts of legal documentation, and draft and review annual reports

There are already some real-world examples of these applications, Jyoti says. For instance, Viable takes a customized version of GPT-3 from OpenAI and aggregates qualitative data from help desk tickets, surveys, customer relationship management tools and customer reviews to help its teams understand what their customers are thinking.

Using the generative AI tool, Viable says it has been able to increase its accuracy in summarizing customer feedback from 66 to 90 percent.

DIG DEEPER: Learn why IT leaders are embracing value stream management.

Jasper, an AI-powered content creation tool, uses a marketing-focused version of GPT-3 to produce blogs, social media posts, web copy, sales and customer-focused content.

And Microsoft, whose Bing search engine generated much interest recently for its embrace of conversational AI, revealed after the new product’s release that it relies on GPT-4. The technology operates using a network of optimized NVIDIA GPUs.

Other firms use the tools to generate digital twins in their design and simulation of new or improved products, Jyoti says.

Photo credit: BlackJack3D / Getty Images. 

How Can Businesses Use Generative AI and Tools Like ChatGPT?

The most important thing for businesses to understand is that they should treat a tool like ChatGPT as a co-pilot or collaborator, says Jyoti.

“It will help with operational efficiency,” she says. “It will help with personal productivity, writing assistance and knowledge discovery.”

Companies can upload their knowledge bases from a variety of sources — PowerPoint presentations, SharePoint drives, websites and other locations — and then teach the models to provide training materials or answers to new or existing employees’ questions.

For example, a financial adviser cannot possibly remember all of the clients a firm has dealt with over the years. But he or she might be able to query a chat tool trained on that material to get an answer to a question about a new case.

“So it’s not just about efficiency, it’s also about scenarios that humans cannot do,” Jyoti says. “I would say crawl, walk and then run, but be cautious.”

These tools likely make the most sense right now as internal-facing solutions and then later in customer-facing applications. And the output from tools like ChatGPT should not be taken as gospel. For example, the financial adviser should apply knowledge and common sense on top of an AI-generated recommendation before he or she advises a client.'

PLAN AHEAD: Understand why companies are considering more superapps.

What’s Next for GPT?

In addition to recognizing that ChatGPT sometimes provides inaccurate responses, there are other cautions businesses should take, the partners at McKinsey note. For example, filters are not yet tuned to flag inappropriate content, systematic biases are still a concern, and “individual company norms and values aren't reflected” in the tools, they write.

Organizations should identify and pilot use cases that are the low-hanging fruit “to accelerate their efficiency and accelerate their time to value,” Jyoti says.

Companies also have to make sure the appropriate guardrails are in place and that they are using responsible-AI principles, she adds.

Moreover, Jyoti says, organizations cannot simply lean on data science teams to drive this innovation. They must involve compliance officials, line-of-business teams and IT teams to decide on use cases and work in collaboration with technology suppliers.

“You do not have a choice to not do anything, because the market is moving so rapidly,” she says. “But do it with the right due diligence, with the right guardrails, with your trusted partner.”