In today’s episode Jenny talks about the six core tenets of using AI in healthcare marketing, including
- Approach AI results with healthy skepticism.
“Hallucinations” are common with AI and it can be easy to assume the information you are getting is factual. - Do not use AI outputs as-is without reviewing and adjusting.
For example, if you have it generate an internal email, do not just copy, paste, and send. Make sure to read it and confirm it’s formatted (and reads) correctly. - Do not have AI create client deliverables.
You can have it assist with ideas or some content thoughts but the deliverable should still be yours. Use AI to help with a first draft, but then edit, edit, edit! - Be very clear in your ask.
Give AI clear directions, ask it to play a role, use specific language, ask for multiple versions, etc. The more information you give in the prompt, the better the result will likely be. - Ask for sources!
Remember to ask it to give you a confidence level along with its response. You can also ask it to provide sources and citations for the information in its result. - Keep privacy a priority!
Be careful sharing ANY data or information with these tools. Until we have a private space using one of these tools, anything you supply it with could be used for training its model.
Connect with Jenny:
https://www.linkedin.com/in/jennybristow/
Jenny: [00:00:00] Hi friends. Welcome to today’s episode of “We Are, Marketing Happy,” a healthcare marketing podcast. I am your host, Jenny Bristow. I am the CEO and founder of Hedy and Hopp. We are a full service, fully healthcare marketing agency, and we are the company behind this podcast. I’m super excited to be here with you today.
I want to talk about something that is very buzzy in the healthcare marketing space right now and that is AI. A lot of questions, concerns, lots of different ways that you can have an opinion about AI, but a lot of conversations are happening right now at the leadership, as well as the individual contributor level with marketing teams, trying to figure out how, when, and if to incorporate AI into their processes.
[00:01:00] There are lots of resources talking about specific AI tools, best practices for each of those tools. We’re not going to cover that today. Instead, what I want to talk about is six tenets or core approaches to keep in mind as you’re beginning to think about how, when, and if to incorporate AI into your marketing team’s processes.
I’m excited and certainly willing to do future episodes about AI. At Hedy and Hopp, we actually have an internal team that is becoming experts on all the different tools, how to use them appropriately, et cetera. So if you want an episode about a specific tool, about a specific use case in the future, et cetera, holler at me.
But right now we’re going to talk about six core tenets. So for anybody who has begun using an AI tool, played around with it a little bit, you know, there are lots of weird considerations you have to have. They operate a little bit like a search engine, but also [00:02:00] very different from a search engine in their outputs and capabilities.
So as you start talking about, you know, how should we begin leveraging AI to make our internal teams more efficient? The first thing I want you to think about is to approach AI results with heavy skepticism. There’s something in AI results called a hallucinate, a hallucination. And, a hallucination is whenever AI basically makes it up.
They don’t know the answer, but they don’t want to sound dumb, so they make it up. So, the information is not always factual, and it can be very difficult to actually discern when this information is a hallucination. So, for example some of the testing that we have done and work that we’ve done, they’ll actually cite specific court cases with specific numbers.
It looks very real until you begin looking for a source, and it doesn’t exist because it’s a completely made up court case. So the first thing is to approach it with healthy skepticism. [00:03:00] Start by assuming the information AI is providing to you is not accurate versus assuming it’s accurate and moving forward with that information.
Number two is do not use the AI output as is without reviewing and adjusting. This should be super obvious based off of tenet number one, but if you have it generate an internal email, for example, don’t just copy, paste, and send. You need to read it thoroughly, make sure that it communicates your points effectively, doesn’t use vocabulary that you wouldn’t use normally.
And another issue is that AI often steals content. There’s lots of copyright infringement lawsuits happening right now. And so it’s a form of plagiarism often in using AI generated output and using it as your own online. So, what do you do instead? Well, tenant number three is don’t have it create deliverables.
If you’re working and saying having to do some social media copy and you wanted to create different versions for all the different [00:04:00] social media platforms based off of the word count. Awesome. That can be a V1, but then take it and edit, edit, edit. It can do some of the groundwork for you. And that’s a great way to do it.
And we may come a day where we pass this and we’re able to use the specific output, but there’s just too many lawsuits and things at play. We’re using the content as is for internal or external purposes. Is not a good idea. So again, do not have it use do not have it create specific deliverables internal or external facing.
Number four is be very clear in your ask. So asking whatever AI tool you’re using, giving it clear directions. Asking it to play a role, such as a world class healthcare marketer, for example, using specific language, asking for multiple versions, et cetera, the more information that you give it in the prompt, the clearer and better that your results likely will be.
A study was actually published. I’ll see if I [00:05:00] can find it to be able to reference it in the show notes, but it actually said, if you’re polite to AI, it actually is more willing to help you and gives you much more comprehensive results than if you do not say things like, please. And thank you. So being very clear in your ask, communicating to it, like you would a human, so being appreciative, asking for the level of comprehensiveness that you need for the results, et cetera, are all super important as you’re beginning to test all these different platforms.
Number five, ask for sources. Some platforms like Perplexity do this really well where if you ask it a question, let’s say for example if I was working on a campaign, a patient acquisition campaign for pediatric Orthopedics in the St. Louis area, and I wanted to understand the population within certain zip codes of people under the age of 18.
I could ask it, but then also ask it to cite the source like where did you get the information and that will help provide comfort that it’s not a hallucination and made up data and you can click through and [00:06:00] actually see the source quote the source, make sure the data and information is accurate. And so it can be helpful for that upstream research, but again, always ask for sources.
And then the final one, and as your neighborhood healthcare privacy pros, I have to mention this one is keep privacy a priority. Anything that you enter into these AI platforms, can be used for training their model. So for example, you never ever want to upload anything that is proprietary information, anything that is PHI, anything that you wouldn’t necessarily just like, you know, share on a billboard with an AI platform because that essentially is what it could be doing in the future, it could be incorporating it.
So for example, to test this theory we’re using chat GPT and I started using the platform, a paid version, to brainstorm content ideas for Hedy and Hopp. For the, “We Are, Marketing Happy” podcast, trying to get sources to make our information more comprehensive.
[00:07:00] And over time, we noticed the results it’s giving us is actually leveraging the Hedy and Hopp branding. So it’s using our own terminology. It’s using the way that we talk and the way that we phrase sentences and structure sentences. And so it very much takes the information that you share and reuses it.
So do not put anything in it, unless you have a private instance on a server that you can control and your IT team has given you the stamp of approval. So again, what are the six tenants to AI privacy? Number one, approach it with healthy skepticism. Assume some of what it’s telling you is a hallucination versus assuming everything is factual.
Number two, do not use the output as is without reviewing and adjusting. Number three, don’t have it create deliverables internal or external. Number four, be clear in your ask. Give it clear directions. Be polite, be comprehensive and specific in what you’re wanting it to do. Number five, ask for sources, tell it to cite sources and give you a link.
So you can check [00:08:00] fact check the information and number six, keep privacy a priority. I’m going to end this episode on a fun little comment. I actually did a LinkedIn post last week that a lot of folks got a good chuckle at. My oldest son is a freshman in high school and he missed an in-person event because he was doubly committed that night for afterschool sports commitments.
He had to write a five page paper. He was very proud when he told me he used AI to write the paper. And I was very clear to him that your teacher is immediately going to know it is AI written. He said, nah, mom, nah, she won’t know. He immediately received an email back because the homework submission tool had an AI tool that really just was able to read the content and identify quickly if it was AI generated or not.
My kids put their teachers emails into these tools to see if the emails and notes home are generated by AI and they often are. Those tools are not a hundred percent accurate, but just know [00:09:00] that from a credibility perspective, the last thing you want to do is kind of shift your mentality where everything you do is going to be generated by AI, because that is noticeable and people can tell.
When the format structure, the terminology vocabulary you’re using just doesn’t match the way that you normally talk. So, word to the wise, be a little careful with it. But there are ways to incorporate it like we talked about that you can, but do not copy and paste. So thank you so much for joining us on today’s episode.
Hopefully these six tenets gave you some ideas about ways to talk about AI with your team, brainstorm ways to use it safely and some things not to do. Again, if you have a specific topic within the AI realm that you want us to cover, shoot me a note. We’re happy to dig into more specifics. Things are changing really fast.
So we’ve kind of avoided platform specific conversations up until now, just because by the time it publishes, it may be out of date, but we are very happy to dig in and have that conversation with you. So thank you so much for tuning into this week’s episode of [00:10:00] “We Are, Marketing Happy.” We’ll see you in the future.
Take care.
