In the second episode of our three-part AI series, Jenny builds on the understanding of AI by covering the six tenets for using AI at your healthcare organization. While AI offers exciting opportunities, it also comes with challenges that require caution and strategy.
The six tenets covered:
1. Approach AI results with heavy skepticism: Even tools like ChatGPT can provide inaccurate information, with hallucinations still occurring about 30% of the time.
2. Do not use AI outputs as-is: AI-generated content should always be reviewed and refined to avoid sounding impersonal or inaccurate.
3. Do not have AI create marketing deliverables: Relying on AI for final marketing assets can be a liability due to inaccuracies, bias, copyright concerns and can cause potential harm to your brand.
4. Be clear in your ask: Use structured prompting techniques to get the best results.
5. Ask for sources: Always ask AI tools for sources and confidence levels to verify their output.
6. Keep privacy a priority: Treat any shared information as though it could appear on a public billboard to ensure confidentiality and compliance.
Connect with Jenny:
•Email: jenny@hedyandhopp.com
•LinkedIn: https://www.linkedin.com/in/jennybristow/
If you enjoyed this episode we’d love to hear your feedback! Please consider leaving us a review on your preferred listening platform and sharing it with others.
Jenny: [00:00:00] Hi, welcome to today’s episode of We Are, Marketing Happy, a healthcare marketing podcast. My name is Jenny Bristow. I am the host of this fabulous podcast. I’m also the CEO and founder at Hedy & Hopp. We’re a full-service, fully healthcare marketing agency. And I’m very excited today to bring you episode two of our three-part series of AI 101 for healthcare marketers.
In our first episode, we talked about how to understand AI if a person presents it to you. Trying to understand and categorize the different kinds of AI at a foundational level to give you a better understanding. And the six core platforms that are used most heavily today
What we’re going to do in episode two, is we’re actually going to talk about the core tenants, the 6 tenants of using AI in healthcare marketing. And we’re going to talk a little bit about how to get buy-in within your internal organization about using these tools. So let’s get started.
[00:01:00] Six tenants for using AI in healthcare marketing. The first tenant is approach AI results with very heavy skepticism. Even ChatGPT knows it lies sometimes. So it’s called a hallucination. All of these platforms are built wanting to please the user, which is you. So if you ask these platforms are questioned that they don’t know the answer to very often they will lie to you.
They will come up with an answer. So you have to really approach all of them with really heavy skepticism. A great example is at Hedy & Hopp, again, we’re a full-service agency. So we do everything from marketing strategy, messaging, persona work, to activation. So big strategy and activation for campaigns to drive patient volumes.
A great example of this is on the privacy side we were digging in and helping one of our new clients be compliant. They were in a state that we hadn’t worked with before. So, before we called our attorney that we have on retainer for all of our privacy work, we thought, hey, let’s see [00:02:00] if these platforms can give us a good rundown of the current privacy laws in that state.
No, the platforms lied. Specifically, ChatGPT lied. It actually made up a court case, including the name of a court case. So had we not been super diligent and actually fact-checking the result we would have given wildly inaccurate information and it does this for all sorts of questions. It could do it if you ask it a statistic, if you ask it any information that it cannot easily access, even the versions that have ready access to go onto the Internet to search. If they can’t find the answer quickly, they will lie. Now the percentage of times hallucinations happen are drastically dropping. Every time that these models are updated, hallucination rates decrease. I saw a recent study that for in particular, from the prior model to this model, hallucinations dropped [00:03:00] from it was like a percentage in the fourites down to in the thirties. And you may say, Jenny, but wait in the thirties is still super high. And yes, it is, which is why we’ve got to fact-check this stuff.
So again, number one was approach AI results with very heavy skepticism. Number two, don’t use AI output. As is it’s not great. It reads like it’s AI. So there will be a time and place where AI models are sophisticated enough where they can do full content creation and copywriting for you. But it’s just not there yet today.
We tested all of the current platforms, but the same prompts and have been doing that over the last year to kind of understand the way that they’re maturing and the way that the models are shifting Tested different prompts, all of these things. You do not want to make any of your users feel as though AI [00:04:00] is talking to them.
So we’ll talk about some ways that you can use AI to compliment your content marketing program through thought partnership, brainstorming, assumption checking, but we’re not going to use the output as is. And number three is do not have AI create marketing deliverables. So a couple of reasons, but the biggest takeaway is it’s a direct liability to your employer or your brand.
There’s four reasons why. First legality. I mentioned this in the first episode when I went through all of the different tools, but ChatGPT is actively being sued because they trained the model using a lot of content where the creator of the content was not given. They did not give permission and they were not compensated.
So, imagine if ChatGPT, you know, next year ends up settling this lawsuit or going all the way through the courts and then being fined, say 2 billion. How do we know that’s not going to roll downstream? And [00:05:00] anybody who used any some certain outputs of ChatGPT are going to be financially liable.
We don’t know that. And so we have to be really careful and understand that using it exactly as is to create a marketing deliverable can be a legal liability. We also are going to bring up hallucinations again, accuracy. We don’t know that it’s accurate, right? We’re not going to trust it more than we trust our own ability to create content or strategy work.
And so that’s important. Bias. AI tools can learn bias from the data and training that it is interacting with. So, you want to make sure that you’re really careful in the way that you prompt, even the words that you use. Maybe creating bias and the results that you’re receiving. So we have to be really aware of that and not using deliverables as is to publish content out on behalf of our brand can protect you from that.
And then fourth attribution and authoring [00:06:00] along with the aforementioned legal issues. If these tools are used to create things that are put into production, there could be an issue. For example, let’s say you’re doing a brand refresh and you’re wanting to come up with a specific tagline. You ask Gemini, for example, to come up with a variety of taglines for you and one feels so good.
It is so awesome. And you launch it. A week later, you find out that’s one of your top competitor’s taglines. No wonder it sounded so good. So again, you cannot take things at surface value. So we’re not going to have it create marketing deliverables for us. The fourth tenant is to be very clear in your ask.
So there’s actually a prompting structure called Risen R I S E N that my friend, Chris Boyd forwarded along to me. And cause I had just kind of been creating all of these prompting structures myself. And he goes, you know, there’s a thing called Risen, right? Like, Oh, thanks, Chris. Yeah, I didn’t. I appreciate you.
But whenever you’re actually creating [00:07:00] a prompt or typing in something to ask the AI model to do something for you, it’s called a prompt. You want to give it clear directions by asking it to play a role, giving it context using specific language, asking for multiple versions, asking for it to cite sources.
Sometimes a prompt or request can be two pages long. So prompting is truly an art form today. Now, will prompting be important in this time in 2025? Probably not. I think back to when again, the Internet was first coming online in the late mid nineties, late nineties. I taught myself. I was in 7th grade in 1997.
I taught myself how to hand-code websites using HTML and JavaScript, and I started a web development business and I built lots of websites in my local community for businesses, schools, and churches. It was really important to know HTML and other coding languages at that point. Today, you can stand up a website using Squarespace and [00:08:00] not have a, any idea about what the code looks like on the backend.
Now, is there still a profession where understanding coding is super important. Of course there is, but again, we’re talking about like the average marketer. And if you need that skillset, I strongly still recommend that if you’re interested in becoming proficient, you should understand prompting and prompt engineering.
There are a couple of different sources. I would recommend Coursera has some great courses around prompt engineering specialization. Udemy has a session called Master AI with Prompt Engineering. It’s 65 bucks. So super affordable. And then LinkedIn learning has lots of new sources. I do know that they run specials from time to time.
If you’re a new user, maybe your employer already has the ability for you to do LinkedIn learning courses. Check it out. Otherwise, they still are really affordable to go through them. It is really important to understand it again now and I still think it’s going to be [00:09:00] important in the future as these models get better and better at kind of like guiding the user where we need to go.
We may not need to know exactly what a prompt should look like at that level of granularity, but it also still helps your brain understand what’s happening behind the scenes. So it’s still a great tool or a good thing to learn. Number 5. This is why I love Perplexity, but also why I think maybe Perplexity may not be around much longer because you can do this for all the models.
You just have to remember always tell within your prompt for the platform to cite its sources. So, you can ask the platform to give you its confidence level and its answer, and then also ask it to provide sources and citations for any information it included the results. So again, Perplexity does this automatically, but all the other platforms, you can just include that at the end of the prompt, and it’s really fabulous.
And then number six, remember, keep privacy a priority. We absolutely have to keep remembering that. Any [00:10:00] information you put within these platforms, imagine it going on a billboard outside of your organization. And if you’re comfortable with that. By all means, move forward with it. The last thing I want to touch on is building consensus within your organization.
So one of the reasons that we created these six tenents is to give our clients and then also just people in the healthcare marketing community, a starting point to have these internal conversations. Just like in the privacy world, if you go to your legal and compliance teams and say, hey, here’s what we’ve been doing with marketing, here’s the diligence we’ve been doing. Here’s the research and the education we’ve been doing. Let’s have a conversation to get aligned. It’s a much more mutually respectful situation than if you go to them and say, hey, will you do all this privacy stuff for me? You know, so let’s treat AI the same way.
Let’s do the legwork. Let’s understand some initial starting points of how we may want to use it within our organization. Let’s help come up with some rules of what we want to see [00:11:00] happen within our marketing team and the usage within our team and organization. And feel free to use these six tenents as a starting point, add to it, modify it, whatever you need to do.
But these are some great guardrails as you’re thinking about rolling it out and implementing it across your team. So thank you so much for tuning in today. I hope episode two of our three-part series of AI 101 for healthcare marketers was helpful. Next week we’re going to cover some actual prompts.
We’re going to dig in and actually look at some specific queries you can type in with some real use cases. So when you’re listening next week, be sure to have me in one ear with your laptop pulled up in front of you cause it’s meant to be an interactive session where you can get hands-on and really build your confidence at starting to use some of these platforms. W
ith that, have a fabulous rest of your day and we can’t wait to see you next week on that episode of We Are, Marketing Happy. Cheers.
