Go to Top

Putting AI To Work, Intelligently

detail of a person using a smartphone and a laptop.

Business owners and executives may feel pressure to incorporate artificial intelligence into their workflows. But given how quickly ChatGPT and its competitors arrived on the scene, not everyone will have an innate sense of the best way to use these tools.

While often casually referred to as “AI,” OpenAI’s ChatGPT, Google’s Gemini and Microsoft’s Copilot are more accurately “large language models,” or LLMs. (More accurately, these three examples are chatbots powered by LLMs, but in the context of this article, the distinction is not significant.) An LLM is an artificial neural network trained on large amounts of existing text. While the nuts and bolts are beyond the scope of this article, think of it as an exponentially more powerful version of the technology behind the autocomplete feature in your email or text messaging app of choice. One of the main applications of these LLMs is text generation, which falls under the umbrella of “generative AI.” This category also includes tools that can generate images, videos or other sorts of data. Much of what I’ll discuss in this article about text generation could broadly apply to business uses for image or video generation too.

Reasonable people can disagree about the pros and cons of LLMs as they currently exist. You have likely encountered both boosters and detractors of these new tools, in the media and possibly in your own social circle. If you are a business owner or senior manager trying to best determine how to use ChatGPT or similar resources in your work, it is smart to take the time to learn more about the ethical and practical issues that are in play. With this information, you can weigh the potential benefits and drawbacks as you implement your tool of choice.

While LLMs are rapidly evolving, there are already ways your business can benefit if you decide to explore them. Tools like ChatGPT, Gemini and Copilot can become valuable assets. And many existing software packages, such as Adobe and Google Suite, have begun to incorporate AI features. Meta, the parent company of Facebook and Instagram, has already begun rolling out its Meta AI to its products and made it available as a standalone website. Some especially large enterprises may even look into building their own custom LLM, though in this article, I will assume you are using an off-the-shelf preexisting tool.

Deciding which generative AI product is right for you will mean first deciding how you plan to use it. I’ll discuss some potential applications in this article, but knowing your goals can help you decide which tool fits your needs. You should also consider your price point. Many LLMs offer some tier of access for free. However, as with any free software or application, you will need to pay close attention to privacy guarantees (or the lack thereof), and educate yourself on what the product can or cannot offer.

Alternatively, you may already pay for enterprise software that incorporates generative AI features. Adobe is rolling out AI tools to software offerings including Photoshop and Acrobat. Apple has reportedly talked to both OpenAI and Google about potentially licensing their respective LLMs to add AI-powered features to its upcoming iOS 18 operating system. Other companies aren’t automatically adding on AI, but offer it as an upgrade. For example, Google currently offers its Gemini as an add-on for existing Enterprise customers, and Microsoft offers Copilot much the same way. Both companies also offer a more limited version of their respective tools for free. If you’re already using — and paying for — software that works for you, explore whether it has incorporated LLM technology or plans to in the future.

Some businesses will want more robust functionality than free or built-in tools can offer. OpenAI offers ChatGPT for free, but its more advanced GPT 4 requires an individual subscription. The company offers higher tier subscriptions for more advanced analytics and certain kinds of customization. Anthropic’s Claude also offers free and paid versions. By design, these tools are broader than those built to integrate with existing applications. The current offerings expand well beyond the scope of this article, and more arrive every day. When you look for the right fit for you, take your time and be willing to think beyond the tools that have made the most headlines.

When you consider how to use LLMs, or any other generative AI, start by identifying repetitive tasks that don’t require significant creativity or human expertise. Some examples may be summarizing an existing report, drafting form emails, identifying trends in data, transcribing voice or video, or suggesting strategies based on information you provide. Offloading these tasks to AI frees up your time and mental energy for more strategic work. In all these cases, the AI’s output can serve as a starting place for you or your employees. Form emails may only need minor tweaks, while brainstormed suggestions will require critical thought from human reviewers.

Beyond streamlining workflows, AI tools like LLMs can also be useful for boosting creativity and innovation. If you’re stuck on an idea, ChatGPT can be a valuable brainstorming partner. For example, imagine a team member who has developed a promising concept, but who is struggling to articulate it or explore its different angles. ChatGPT can help refine that staff member’s thoughts, suggest unexpected approaches, or even generate variations of the idea to spark further exploration. In this example, ChatGPT isn’t replacing your team; it supports them as they try to bring their ideas to life. Ideally, team members can then bring more fully developed or refined ideas to the table, saving everyone some time.

Given the companies that are developing LLMs, one application that is likely to expand is beefed up search capabilities. We are likely to see existing search engines incorporate so-called semantic search capabilities — how search engines parse human phrasing — rapidly increase. For now, LLMs may be more helpful when trying to brainstorm or refine search queries that you can then enter into a traditional search engine. Because LLMs are trained to respond to natural language, they may be able to help you figure out how to approach a research question that has you stuck.

For all of these uses, bear in mind that the better your prompt is, the better the output will be. Don’t underestimate the power of a good prompt with an LLM. Sure, you can simply ask ChatGPT to “translate this paragraph.” Your results may come back in five different languages, which may or may not have been the ones you wanted. And as we all know, translation can be tricky even for a human fluent in both the original and target languages. Message and tone can easily get lost. A stronger prompt would be something like: “Translate this paragraph from English to Spanish, maintaining a formal and professional tone.” This instruction specifies the source and target languages, and instructs ChatGPT to prioritize the tone when translating. Even with a well-crafted prompt like this, I would personally still rely on a native Spanish speaker to double-check the translation before relying solely on ChatGPT’s judgment.

This brings me to my final suggestion, which is possibly the most important: Build in time to fact check any materials generated with AI. A good LLM may seem magical, but it generates text that is likely, not necessarily text that is accurate. In 2023, attorney Steven A. Schwartz made headlines after he and his partner filed a legal brief generated by ChatGPT. The brief cited a variety of judicial opinions and legal sources that turned out not to exist. Schwartz said that he had not understood that ChatGPT could fabricate cases.

LLMs can and will sometimes offer information that is not accurate. Before any machine-generated materials leave your hands, make sure you can externally verify that everything is correct. This fact-checking is an essential quality control step. In the translation example I mentioned earlier, this is why it is critical to have someone who understands the target language review your results. Yes, it will take time. But in many cases, reviewing and checking will still be quicker than starting the entire task from scratch. And, in cases where it isn’t quicker, you may have identified a task that isn’t quite right for an LLM.

Think of generative AI as having a skilled helper on your team who tackles the time-consuming tasks while you focus on the bigger picture. Just like a supervisor overseeing less-experienced employees, you are still responsible for overseeing quality and ensuring everything runs smoothly. There are tasks you wouldn’t give a bright but inexperienced intern, and there are tasks you should not give to ChatGPT. But if you identify areas where these tools can add value, refine your ability to create strong prompts and check the output carefully, generative AI can be a helpful tool for you and your team.