Skip to main content
Artificial IntelligenceStrategyTechnology

Don’t get caught in the headlights: a wiser path to ChatGPT implementation

By July 20, 2023No Comments
Path to ChatGPT implementation banner

Quick summary: While others waste time and resources chasing the perfect use case for ChatGPT, your organization can be laying the foundation for long-term success in getting the most out of generative AI.

 

In November 2022, the holidays came early for AI aficionados, as OpenAI launched its generative AI tool ChatGPT to the public. What ensued was a frenzy that threatened to redefine the word “viral.” In its first two months, the application achieved 100 million monthly active users—a feat that took TikTok nine months to achieve.

 

Business users were not immune from the initial tidal wave of enthusiasm, and LinkedIn feeds began to overflow with glowing accounts of ChatGPT revolutionizing business processes. Thus began this year’s equivalent of the 1849 Gold Rush as businesses launched themselves into a quest to find the ideal use case for generative AI.

 

Then came the inevitable march of the “experts.” Just for laughs, I did a Google search for “ChatGPT expert” and was met with 58 million results, including no fewer than four sponsored links at the top of the page.

 

While it’s easy to get swept up in ChatGPT mania, following the crowd can be a recipe for a massive waste of time and resources, if not regulatory disasters and/or PR calamities. Recently a New York attorney allegedly decided to cut some corners and use ChatGPT to do his legal research for a case. The result was a slick 10-page brief citing similar cases in the past, with just one problem: every cited case was completely fabricated. The judge was not amused.

 

In this article, I’ll propose a more prudent approach to generative AI that heightens your chances of maximizing its benefits.

Nothing new under the sun

Contrary to a popular belief, OpenAI did not invent either generative artificial intelligence or the large language models (LLMs) that give ChatGPT its conversational capabilities. Here at Logic20/20, we’ve been using LLMs in our work for clients since 2019.

 

What OpenAI has done with ChatGPT (as did Google with Bard) is build these technologies into a widely accessible, user-friendly tool that average people around the world can leverage for a wide variety of uses. (Did I mention it’s also free to use?) Neither generative AI nor LLMs have ever been implemented on such a massive scale, and we still don’t know what the end point of this massive upscaling will look like.

 

Build your foundation first

Rushing to plug generative AI into any conceivable use case without building a solid foundation is the equivalent of jumping into an arm-wrestling competition without first building the required muscle.

 

So, how do you build this “muscle?”

 

Get the lay of the land

Start by learning the ins and outs of how generative AI works and putting in some hours of practice on low-risk prompts that do not involve sensitive information. These might include drafting a quick answer to a question, summarizing a document, or writing a simple brief.

 

You’ll quickly begin to see that ChatGPT and its cohorts are currently brilliant at some tasks and outrageously bad at others, and you’ll discover the cases in between that can yield solid results with the right query wording and/or follow-up questions. Keep abreast of new ways of using the technology and new frameworks, such as LangChain today, which may be able to accelerate our use of the technology.

 

Know the risks—all of them

You’ll also want to gain a solid understanding of the risks involved in applying generative AI to business use cases, data governance, data privacy practices, and compliance with regulations (which are trying to quickly evolve, too). There’s also the risk of the phenomenon known as “AI hallucination,” in which an AI platform, upon encountering a question it can’t answer, will simply make up a response, as it did for the unfortunate lawyer we referenced above.

 

Take advantage of what’s already been done

Finally, you can “stand on the shoulders of others”: Use open-source code to release low-risk applications (like Microsoft’s search demo code) and get business users to interact with the tools.

 

Setting up for generative AI success

As tempting as it may be to launch on a quest for the perfect ChatGPT use case, the more measured approach we described above is far more likely to set you up for success. Once your team reaches the desired level of comfort and proficiency in using generative AI for simple, low-risk tasks, you can feel confident expanding into more advanced applications.

 

We do believe that ChatGPT is a transformative platform with the potential to disrupt a host of current business practices. At the same time, we advocate tuning out the noise, learning the game from the ground up, and using this knowledge and expertise to achieve a higher level of generative AI mastery.

 

Nick Maddock

Nick Maddock, Managing Director of Strategy & Operations at Logic20/20, is responsible for consulting innovation and the close partnership between consulting and sales.

 

Logic20/20 is a Solution Provider member of Utility Analytics Institute. Check out UAI Communities and become a member to join the discussions at Utility Analytics Institute.