From Concept to Clarity: Why Communication Skills Are an AI Superpower

One of the most revolutionary aspects of generative AI (GenAI) is that users can enter prompts in plain language. Most of the world speaks at least one of the more than 50 languages ChatGPT supports, for example. But that doesn’t mean everyone is equally able to make the most of this technology. Effective GenAI use requires thoughtful planning and razor-sharp communication skills. 

Context is essential in any conversation, but it’s especially important when working with GenAI, because these tools don’t make allowances for vague and ambiguous instruction. “There’s some early evidence that coders are actually worse at using AI because they think it works in a rational kind of way,” said Ethan Mollick, associate professor of management at the University of Pennsylvania’s Wharton School and co-director of the Generative AI Lab at Wharton, during an interview on the Stanford Graduate School of Business’ Grit & Growth podcast. “And it doesn’t.” 

AI “rises to the level of your game,” said Stephen McConnell, an adjunct instructor at New York University, where he developed the Communications and AI certificate program. “You have to bring your ‘A’ game to the encounter with AI. Converse with it on a level that you converse with your colleague.”

Becoming a Prompt Engineer 

The discipline of optimizing your input into a GenAI tool is called prompt engineering. According to IBM, “Prompt engineering is the process of writing, refining, and optimizing inputs to encourage generative AI systems to create specific, high-quality outputs.”

The problem is that most people don’t provide enough context. Let’s say you walked up to a member of your team and said, “Write an email announcing the creation of a new department.” What are the chances they would produce what you needed? Chances are they’d have some follow-up questions. GenAI tools won’t ask for clarification unless prompted to do so, so you need to anticipate the information the tools need to fulfill your request. 

The clarification your team member would ask for is the same information GenAI needs: 

• Objective. What is it you want AI’s help producing or doing? What is your endgame? 

• Context. Who is your audience? What is your desired outcome? 

• Details. What key information should be included?  

• Perspective. How could different perspectives lead to different responses? 

“We have to know our game,” said McConnell. “The more we know our game, the better our use of this technology will be.” 

Before you go to the AI model, you have to think through the ask. In the example of the email, this would include: Who is the email from? Who is it being distributed to? How many people are going to be in the new department? What will their roles, functions, and responsibilities be? What is the name of the department, and why is it being formed? And give GenAI a role: You are a professional communicator from “X” company (provide a link to your company website) who is adept at composing internal corporate communications. 

Then, go to the model and type in what you wrote down, but in full sentence form. “What you’re doing by giving it an industry and specialty is tapping into its vast sea of training data to hone in on certain parts of its thinking,” said McConnell.  

You already have the playbook to make GenAI work effectively for you. The skills you need are some of the basics of communication, including writing, critical thinking, and problem-solving. You know how to think through what details and nuances need to be included in any piece of communication you produce. So when you turn to GenAI, use those same tenets: Be clear about your goal, use easy-to-understand language, write in complete sentences, provide context, and share samples. 

AI models are “trained on human language, and they’re refined on human language,” said Mollick. “And it turns out that they respond best to human speech. … Telling it to give a task like a person often gets you where you need to go.” 

Aiming for 80%

Even a well-thought-out prompt won’t produce perfect results. However, “it takes you from zero to like 80%,” said Chris Ortega, CEO of Fresh FP&A in McCordsville, Ind. “You don’t have to create things from scratch anymore.” 

“At the end of the day, you might be engaging in a 20-to-30-minute dialogue with the model, but that’s much better than the two to three hours that project would have taken on your own,” said McConnell. 

Ortega relayed a poignant example of this in a story about helping his friend produce an intricate accounting document, an ASC 606, for a client. His friend had spent four hours the night before working on it and had yet to complete it. Ortega started asking him questions about it to get more details, and as his friend answered, Ortega constructed a detailed prompt on his phone and plugged it into ChatGPT. 

You have a manufacturing client, and you are tasked with creating a two-page memo for sales and the CEO that’s clear and concise and easy to understand. Source Ernest & Young, Price Waterhouse Cooper, Deloitte, PWC, and other authoritative information, and include that in your memo summary. Make sure there are sections in bullet points. Have the bullet points be less than three to summarize everything, and also give me a one-page summary in PowerPoint that simply explains ASC 606 to a sales and non-finance professional. 

“He gave me everything I needed,” said Ortega. “We added some customizations, and then the final product was done. And he’s sitting there looking at this, and he says to me, ‘Chris, I don’t know what you did, but if I had known I could have done this, it could have saved me at least three hours last night.’ 

“I asked him, if you’d had those three hours, what would you have done? He said, ‘I would have been able to go to my son’s baseball game.’ ”

The more details you can give GenAI, the better the end product will be. “You have to be very intentional with it. You have to really know what you want, and give it high-quality data to get what you want,” said McConnell. He recommended uploading documents, when available, that provide the information it needs; giving examples such as in the prompt above; and using terms and vocabulary from your field. “It will understand it,” he said.  

Prompting doesn’t have to be a single-input affair. If you want to add an extra layer of understanding, you can end a prompt by telling the GenAI to ask follow-up questions as needed. Or if you have more data for it to use, you can tell it to wait to produce the output. McConnell uses the following statement: “I want to give you some more detailed information. Please do not proceed until I give you that information.”

Be prepared to put in the time to refine the product the GenAI gives you, just as you would a colleague. For example, maybe the tone isn’t right, so you tell it to make the tone more formal or more friendly. “Be willing to have a conversation with the AI [model] that lasts a little bit longer than you think it should,” said McConnell. “You’re giving it feedback the same way you would a colleague.”   

“There are a lot of reasons people stop using AI,” said Mollick. “It’s weird, it freaks them out, it gives them bad answers initially, it doesn’t feel that profound. You need to push through. There is a point of expertise with this where you start to get what it does and what it doesn’t, where you need a cliched result when you might get something interesting.”

We also need to be sure to scrutinize what AI produces for both truth and bias. Regarding truth, when AI doesn’t know the answer to something, it doesn’t tell you that. Instead, it makes up an answer, referred to as a hallucination.   

GenAI has been trained on data produced by humans, and humans have biases. Therefore, it’s important to check the copy it produces for any biases. You may also need to provide it with cultural references or nuance, depending on who your audience is. 

You will also want to adhere to the “use it or lose it” mentality when it comes to using GenAI. Most experts agree that as with any skill, interacting with AI has to be practiced regularly to achieve greater and greater refinement. “The No. 1 thing I told my team was, ‘You have to use this every day,’ ” said Ortega. “It’s like a foreign language: If you speak it every day, you’re going to get better and better at it. If you learn it and don’t speak it for a month, you lose some of the refinement.” 

GenAI Can Improve Human-to-Human Communication

Communication remains a hurdle for many organizations. Out of the 31 million employees interviewed for Gallup’s State of the Global Workplace 2022 Report, only 13% reported that their leadership was practicing effective communication.  

“I continually encounter senior executives who are poor communicators,” leadership coach Abiola Salami wrote in a 2023 article for Forbes. “They assume that talking often is enough, but the feedback they ultimately receive shows that they rarely communicate in a meaningful way.”

Writing (or speaking) prompts for GenAI helps you practice intentional communication, which is conscious, deliberate. and explicit. And when leaders communicate intentionally, the entire organization benefits.  

According to the Center for BrainHealth at the University of Texas at Dallas, “Intentional communication not only makes your message clearer; a top-down approach to sharing ideas encourages others to stay curious and reciprocate effective communications. [It also] creates less room for emotional bias and errors in thinking.”

The only person who will get frustrated if you don’t write an effective prompt for GenAI is you, making it a great testing ground. By asking GenAI to produce exactly what you instruct it to, seeing what the result is, and scrutinizing that result, you will elevate your ability to communicate instructions and feedback to your team and your colleagues with intention.