We’re so excited to have Susan Beard from NNU speaking at Brand Discovery 2024! Susan will be going in-depth on how to set up AI strategies and policies for your company, and why you need them. She sat down with us for a conversation on a few of the considerations companies need to be aware of.
Join us at Brand Discovery for more!
Susan Beard oversees partnerships & business development in the Center for Professional Development at Northwest Nazarene University. The Center for Professional Development offers both team and online training in leadership and a myriad of other topics for the business community. Susan teaches courses on leadership, networking, and communication. As an active member of the community, she thrives when she helps others connect!
Susan: What percentage of employees in the marketplace already use AI without any permission from their bosses?
I've done it. I don't know if you guys have done it, but I have. I was on Microsoft Bing Copilot today, asking, Hey, if you were going to give a 10 minute talk on imposter syndrome, what would be your top three topics?
Right? I do it all the time. And a lot of employers have no strategy in place for protecting their intellectual property around AI.
Specifically, most people don't realize that if you were to put your company's economic plan into ChatGPT and say, give me a summary of this, those data analytics are now part of ChatGPT.
So any other company that does what SOLV does can go in and say, "tell me how other brand companies approach economic development." And the AI would be able to access the analysis of the data put in, like word count, keywords, etc. to form its replies. While it won't tell you, "SOLV does this", it would pull from the words and phrases that SOLV put in to form a response.
And so there are some AI policies that companies should really have in place:
Susan: I'll give you an example. Last week we were talking about developing a five step marketing strategy. We asked, okay, we're going to do a social media campaign, what five things do we want to talk about?
And so as a brainstorming tool, I asked ChatGPT and Bing Copilot to tell me five reasons that somebody would want to attend NNU.
And it gave me some good information, but the last thing that it said on ChatGPT was, check out the Picasso in the De Gaulle Library, blah, blah, blah.
I think a lot of people don't realize that ChatGPT and Microsoft's Copilot are just pulling information based on what might attract somebody to this. And it's not necessarily true.
So one of the key aspects from a legal perspective is you have to vet the information that you get. You cannot just trust that it's all accurate.
I've seen people ask things like, tell me about the executive leadership at a specific company, and it will actually give the wrong people who work there, or people that worked there 20 years ago. Because it's culling information from the from the internet and it's not making sure that it's still current.
At NNU we have a running joke about who can find the most incorrect information through AI.
And so from a legal perspective, a company needs to think about protecting their proprietary information. They need to think about having an AI policy. They need to think about which tools they will allow employees to use. They need to set a strategy for what it will be used for, and make sure that people are only using it when you they're logged in so the company’s data policies will be in place.
At NNU, we use Microsoft, and when I go into Bing Copilot , it says, “your information is protected”, which means that whatever I put on there, if it's collected in their saved information, it doesn't have identifiable information about me or about NNU.
Matt: Going back to the idea of putting out the wrong information, that’s fairly easy to control when you're in a small organization. Let's use SOLV, for example. If an AI shot out something wrong about what we have, like our locations, we could catch that one super easy.
But if you were at NNU where there's a ton more employees, and maybe some who have never been to the library and don’t know anything about it, they could have just put that information out there without even knowing. And so I'm very curious, how would you think about those policies in terms of scale? Because It's easier to manage at a smaller level than it is at a larger level.
Susan: Right? Companies that are really large, like a Texas Instruments or a state government, generally already have internet and AI policies. Many employees don't read them and don't know. But there are companies out there already that block ChatGPT and Microsoft Copilot. You can’t actually get to them on a work computer.
But in a smaller company, the first thing to do is to have a policy that says if you are using ChatGPT, you have to verify the information before you put it out in the public.
One of the things that I do is say, where did you pull this information from? And then it tells you where it got the information. And then I can say, this looks questionable. Where did you get this? Or is this accurate? And, you know, with theDe Gaulle Library it said, you're right, this is not actually at NNU. I asked, where did you pull this information? And it just said, this information is not accurate. We recommend you do not use it. <Laugh> But it didn't tell me where it came from.
That's one of the key aspects: know what your purpose is, be specific, and then verify your information. That‘s a strategy that most companies should have in place at the very minimum.
And then another strategy is, know what you're going to use AI for. So if you're going to use it in your marketing department, you would want to have some training ahead of time so that people understand prompt management. People should be trained on that and not just go willy-nilly onto AI and start asking things.
A lot of what you get out of ChatGPT is based on what prompt you put in. And so from a policy perspective, you would want your employees to have training ahead of time. I know that training isn't sexy, it isn't very exciting, but having somebody work with your company about building those prompts is going to protect you in the long run.
Susan: Right? And again, that comes back to having an AI prompt management policy.
It doesn't always give you something because a lot of times it's pulling pieces, not a whole thing, right? So it's pulling a little bit from this company and a little bit from that company and this, this over here, which should protect you, right? Because it's not all from one company.
But that's why it's important to ask those prompts to know what the process is.
Susan: I asked ChatGPT to give me a title for a course, and what might be the objectives? And then after it gave me something I liked, I put it in there and I said, can you tell me if anybody else is doing a course like this?
No, nobody's doing it. Then I ask, where did you pull this information from? And it pulls up articles. This article, this article, this article. So now I can cite those sources, right? In my course, I can use those sources and cite them.
But again, it goes back to prompt management, and remembering that if you pull stuff off the internet, it's pulling from millions and millions of sources, and you want to make sure that you are covering your intellectual property and making sure that you're not stealing from someone else.
Prompt management is asking, where did you get this? Do you have any articles that I can cite? Better yet, provide the link so that it doesn't look like you’re making stuff up out of thin air and stealing people's information without giving them credit.
Susan: There aren’t a lot of protections. The first thing that I would tell you is, don't put anything that's proprietary into an open model. If you have a recipe and you're a restaurant, don't stick your secret sauce recipe into ChatGPT. If you've developed a proprietary formula for cleaning solution, do not put that in there because now it belongs to ChatGPT.
There are AI options that you can purchase that will protect your data. I don't know how much I believe that it completely protects your data. I mean, I'm very skeptical by nature. But ChatGPT-4 is a paid subscription and it gives you that little layer of ability to get more refined information, more accurate information.
But to be honest, I would tell somebody that if you have a secret anything, you should not be putting it into an open AI. You should just not do it.
Matt: What do you think a lot of people are worried about when it comes to AI? What industries or jobs do you think may be worried about this whole machine learning model?
Susan: Well, anybody who's in any kind of code development, I think that they would be concerned about code that comes out of ai for that intellectual property piece.
But what we learned and what we've seen even at NNU is, is the industry that's being most impacted is marketing. Our students that are majoring in marketing are having a harder time getting internships because a lot of people are using AI to do their marketing strategies.
Susan: The reality is that AI can pull information, but it can't create information. You know, Matt, you're an artist. There is no way that AI could make your art, right? It can pull from other stuff and create something that you could tell it, I want something that looks similar to this, or these are the colors that I'm interested in with a mountain scape or whatever.
And it can pull information. But it can't actually create something.
It also doesn't have the ability to empathize or feel. It can do your chat box, right? You can use an AI to respond to people that ask, Hey, where's your office located? You can use AI to do repetitive type tasks that have consistency in them, where the same kinds of questions are being asked over and over. But it can't do anything that requires human sympathy.
And so for companies, it's really identifying what can AI do, what's it good at and what is it not good at? It is not good at showing empathy. It doesn't have common sense. If you have a limited data set, it's not going to be able to give you accurate results. It's only going to be able to analyze the data in front of you.
So if you put your data in from a LinkedIn ad, and it pulls that information, it can't take just that one data set and extrapolate a whole normative study on whether or not it was effective.
But when it comes to dealing with a bunch of information, it can easily create a summary for you. You know, here's all this information. Let's say you did, I don't know, 20 interviews with people about a product that you have, and you enter all that information and say, summarize what our focus group thought of this product.
So rather than you inputting all of the information, you can upload it and say, here is the feedback that we got from this focus group, summarize it for me. And it can tell you we did not like the color, we did not like the tone, we loved the graphics. Right? That kind of thing.
Matt: I like your perspective as somebody who works in education on how it's really affecting younger people trying to get into a workforce position, and that especially in marketing, AI doesn't have that empathy or that unique perspective. You need to keep that human piece to it. But right now, since it's making people's jobs a lot faster, why would you hire more people if AI can get it done in half the time?
Susan: One of the things to keep in mind is that if you pull it out of ChatGPT or Bing or whatever, it's been done somewhere before. It may pull different pieces and feel fresh and new, but the reality is whatever it's pulling is already out there or it wouldn't be able to pull it.
So it doesn't really have the opportunity to innovate because it's pulling from things that are already out in the cloud. Right?
If everything that you do is pulled from stuff that's already been done, are you an innovative company?
Just because you're using cutting edge technology doesn't make you innovative. If you're really hoping to be innovative, you need people that are collaborating and talking and using personal experience and, you know, what worked, and I saw this and found it really interesting, and then you can ask ChatGPT, hey, we're thinking about this. What have you seen in this area?
And it can give you some information, but again, it's all stuff that's already out there, right? And if it doesn't give you any results, you know that it's super innovative. I haven't found anything on that <laugh>, right?
Matt: Yeah. I want print that on a t-shirt. Just because you're using innovative technology doesn't mean you're an innovative company. I think that's right.
Susan: Absolutely. And nothing is going to replace that human interaction.
AI is not perfect, and you have to keep tight controls on it to make sure that it doesn't venture outside of the boundaries, because it happens all the time.