You Ask, I Answer: Generative AI and Public Relations, Part 4 of 5

In part four of my five-part interview with Stephanie Moore, you’ll gain practical insights into ensuring the responsible and ethical use of AI in your public relations work. Discover the importance of fostering a company culture that prioritizes ethics and learn effective strategies for disclosing AI use to clients without fear of negative repercussions. We’ll also explore how to train your team effectively, emphasizing the value of comprehensive prompt engineering and a continuous learning approach. Tune in to learn how to confidently integrate AI into your PR workflow and elevate your team’s capabilities.

You Ask, I Answer: Generative AI and Public Relations, Part 4 of 5

Watch this video on YouTube.

Can’t see anything? Watch it on YouTube here.

Listen to the audio here:

Download the MP3 audio here.

Machine-Generated Transcript

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for watching the video.

Christopher Penn: In this episode—this is part four of five of an interview I did with Stephanie Moore of Kent State University about generative AI in public relations.

Stephanie Moore: Makes sense. Kind of in that vein then, what best practices do you follow to ensure the responsible and ethical use of AI? And then maybe you can talk a little bit more about how you disclose the use of AI with your clients. I know you do that because I actually see in your newsletter, you have disclaimers.

Christopher Penn: Yes. So, “responsible use of AI” is kind of a meaningless fluff term. “Ethical use of AI,” again, that goes to—if you want AI to be used ethically, make sure that ethical people are using it. So, try to avoid hiring jerks.

But also, your corporate culture has to be one in which you prioritize high-quality results.

Stephanie Moore: Right.

Christopher Penn: The ugly truth is, most agencies do not. Most agencies prioritize meeting client deadlines, getting things done as efficiently as they can because you’re working 80 hours for a 40-hour paycheck, and client demands are never-ending. And clients have—and, to some degree, understandably—they don’t care what your priorities are. They think that they are the number-one priority all the time.

And so, if you operate in a “my-hair-is-always-on-fire” methodology, then you will make ethically questionable choices. And so, part of ethics is having leadership and client policies and things that say, “We believe in ethics. We believe in doing as little harm as possible. We believe in things like the truth, and we will enforce that with our clients.” And you empower a team member to say to the client, “Nope, you’re not going to get that today, and you’re probably not going to get it tomorrow either. Sorry.”

Or you say, “Oh, you can have that today, but I’m going to add a $40,000 charge on your bill. Do you want to proceed? Yes or no?” And that typically helps to—do some of that.

So, that’s how you ensure ethical anything: is giving people the ability to say no to things. If people don’t feel like they can say no, they’re not going to make ethical choices.

In terms of the disclosure of ethics, of AI—so, there’s two things to keep in mind here. Number one, in some parts of the world, it’s required now, like the EU. You must disclose the use of AI.

Number two, in a study from MIT Sloan last fall, they did a test. They showed content to three groups of people. The first group was just told, “You’re going to rate this content.” They were not told who made it. The second group of people was told, “You’re going to be rating this content. Some of it was made by AI. Some of it was not. You will rate it, but you know that some of it’s machine-made.” The third group of people was told, “You’re going to rate this content. Some of it’s made by AI. Some of it’s not. Each piece of content will contain a disclosure of who made it.”

When consumers consumed the content unlabeled, they preferred AI content over human-led content. When they were told that humans were involved, they boosted human scores, but they did not penalize AI.

And so, one of the key messages from that study is it’s okay to disclose that you’re using AI. You will not incur reputational harm from disclosing that something was machine-made.

How you do it is pretty simple: like, “Hey, this piece here was made by AI.” And it’s important to disclose not just that you use AI but what part of the content was made by AI, or how it was used because we go back to copyright. If I have an AI-generated image on my blog post, but I wrote the human blog post, and I just say “I used AI,” the logical assumption would be that the entire post is made by AI, and therefore someone else can copy it because it’s in the public domain. And then you have to go through a whole lawsuit of trying to enforce your copyright.

If you disclose, “The header image was made by AI, but I wrote this blog post,” you are reasserting your copyright over the portions that are human-led. And therefore, in that disclosure, if someone then copies your blog post, you retain your rights, and you can go sue them for infringing on your copyrights.

So, disclosure is just as much about protecting you as it is helping the end consumer.

Stephanie Moore: Great. That’s great advice. Okay. All right. Moving on. Let’s talk about some pedagogical approaches in terms of teaching AI. So, how do you train your team about how to use AI for strategies and tactics?

Christopher Penn: The biggest thing we do is we have a shared prompt library. And we, basically, you know—it’s, obviously, very, very proprietary—but we have, essentially, a code database because prompts are code. They’re just code written in plain language. And people can pick them up, adjust them, tune them to their specific use cases, or generate new ones.

We also teach people our methodologies. They’re on our website. You can download the PDFs for free of the basic methodology. And then there are advanced methodologies that we, obviously, keep to ourselves. But, say, “Here’s how you develop this set of outcomes based on what you wanted to do.”

A big part of it is helping people understand the principles behind how these machines work and the structure of getting to the output you want. And my co-founder and CEO, Katie Robbert, talks about this a lot: really skillful prompting and use of AI looks an awful lot like good management of people. If you are good at giving direction and delegating and giving measurable outcomes, you will be good at prompting. You just treat it like a junior employee, and you work with it in exactly the same way.

If you’re a bad manager of people, you’re also going to be bad at AI because you’re going to get frustrated. You’re going to give it incredibly short directions. To give you a sense, a prompt in generative AI today, in Google’s Gemini, it can be as long as 2 million tokens. Two million tokens is 1.5 million words. That is two copies of Shakespeare. Two of these can go in a prompt.

So, these people sharing their, “My ultimate ChatGPT prompt,” that’s a paragraph long—like, no. That’s like saying, “Here’s my secret cookbook, and every recipe is one paragraph.” Yeah, you’re not cooking anything with that except your shoe leather. Whereas, if you look at a proper recipe in a cookbook, it’s at least a page, if not two pages, long.

That’s a good benchmark for how long your prompts should be. And we tell people as part of training, if—

Christopher Penn: —your prompts—it depends on the task—but, generally speaking, your prompts should be at least a couple of pages long. Most of my day-to-day prompts that I use are four to six pages long.

Stephanie Moore: Great. So that’s—it reminds me of the “garbage in, garbage out,” right? More context, the better. So, um, any specific challenges that you have with people learning about AI use—

Christopher Penn: —within our organization, it’s—it’s helping people focus and not get overwhelmed because there’s so many tools, toys and tools and gadgets and things that you can play with and stuff like that. That’s a big part, in our organization. Generally, when working and training others—because I do a lot of keynotes, I do a lot of workshops and things—is getting people to think like programmers. Getting them to think in the way that—you know—”how do you skillfully give instructions?” can be a very, very challenging thing to do. Also, getting—

Christopher Penn: —people to overcome their resistance to trying the technology because they have, in many cases, a preformed opinion of the technology, good or bad. And if they’ve had any experience, it’s been novice-level experience, where they have not seen the use cases that make them go, “Holy crap, I want that.”

I’ll give you an example. I was working in the PR department of one of my clients, and they’re like, “Oh, yeah, we’ve used AI a little bit, but it didn’t do such a great job.” I said, “Okay, well, let me show you this.” I—we took a blog post and fed it into a very long, programmatic prompt. And it said, “Okay, you’re going to come up with six pitch ideas for local news outlets.” And it gave a list of the news outlets and the coverage and the types of things they wanted. And it generated some great pitch ideas, like, “This is really incredible.” Like, their eyes were open, and they said, “I understand the technology because I could see the benefit.”

I’ll give you another example. I was going to pitch this one reporter with—relatively recent. And I said, “I want to pitch this reporter.” It was to query something that they won’t have to work very hard with. But I want—

Christopher Penn: —to, obviously, say the things I want. So, I grabbed the reporter’s last 20 articles they wrote about, and I told the generative model, “Learn this reporter’s writing style down to the letter.” And it said, “Here’s my—here’s my pitch. Rewrite my pitch in the reporter’s writing style as if the reporter wrote in it,” reformatted and stuff like that. And I said, “Here, there you go.” And the poor—is like, “This is great. I don’t have to do any editing.” Like, “No, you don’t, because the model did it for you.”

Stephanie Moore: Wow, that’s brilliant. These are great ideas, too, for the classroom. So, thank you. All right, so I’m kind of moving on then from learning. How do you assess the effectiveness of using these tools in meeting communications goals?

Christopher Penn: What are the goals? That’s right. Save money, make money. If you do one of those three things, you’re doing a great job.

Stephanie Moore: Okay. And you kind of already shared examples of that. How about assessing your team’s understanding or proficiency in using AI for strategies and tactics?

Christopher Penn: We don’t have a formal process for that. We have, on the team, as overused a term as it is, we have a continuous learning process. You know, every time someone on the team is like, “How do I do this?” we’ll talk it through and say, “Okay, here’s how you do it.” And then if the person has a good prompt or they have, “Hey, like, this prompt didn’t work to what I thought it would do,” and we troubleshoot it. We troubleshoot, you know, live, or in team meetings. We have multiple get-togethers throughout the week with folks to say, like, “How are you—you know—how are you getting on? What are you stuck on? What things do you need some additional assistance on?”

A really good example that is, we do a lot of—you know, as everyone does—proposals and scopes of work. We have a standard proposal template. We have a standard master services agreement. And so, one of the things that I showed our sales team how to do is, like, “Here’s the call transcript from your last input call, how—here’s—you tell the model, ‘Here’s the scope template, here’s the MSA, here’s the call transcript, write scope, but,’ and—you know—there’s a list of, like, 20 rules, ‘these are the rules for this scope.’” And proposal and scope writing went from taking six to eight hours to, like, 15 minutes—15 minutes for them. From the time a sales call is—input call is done to the time the proposal is going out the door is so much shorter now. It’s great.

If you enjoyed this video, please hit the like button. Subscribe to my channel if you haven’t already. And if you want to know when new videos are available, hit the bell button to be notified as soon as new content is live.

You Ask, I Answer: Generative AI and Public Relations, Part 4 of 5

Machine-Generated Transcript

Comments

Leave a Reply Cancel reply

Pin It on Pinterest