Latest / #ContentChat / 9/28/26 #ContentChat: Turn Your Content Foundations Into A Consistent AI Instruction System

Transcript

9/28/26 #ContentChat: Turn Your Content Foundations Into A Consistent AI Instruction System · #ContentChat

• 0:03
  SPEAKER_00: Good afternoon and welcome to content chat.
• 0:07
  SPEAKER_00: I'm your host, Erica Heald, and today we're going to be talking about what to do after you have all of your content foundations in place.
• 0:16
  SPEAKER_00: Because it's one thing to put together your brand kit for how to you for all the tools that you've created to help people with using AI that still maintains your brand voice.
• 0:29
  SPEAKER_00: But that's not the end all and be all.
• 0:31
  SPEAKER_00: You actually need something to help them when they're engaging with those AI tools so that way they understand which of those foundations they should be putting to use in what manner in order to ensure that they're getting those on on-brand outputs every time.
• 0:50
  SPEAKER_00: So what I ended up creating as a result of the work that I've done over the past few years, while then helping clients put the content foundations in place.
• 1:04
  SPEAKER_00: I created what I'm calling the AI instruction system template.
• 1:08
  SPEAKER_00: And yes, I have it printed out because I want to make sure that I go through all the elements.
• 1:14
  SPEAKER_00: You'll see it's uh significantly more uh comprehensive than just putting together a sentence or two into a prompt, and it's really a different approach.
• 1:26
  SPEAKER_00: It's not looking at prompts as separate one-off things that you do in order to get a result.
• 1:34
  SPEAKER_00: Instead, it looks at having a consistent method for using AI tools across your entire organization and regardless of which platform someone is using.
• 1:46
  SPEAKER_00: Now, this is incredibly important because even if you feel as though everyone is setting up various incendary tools and that they've you know trained those tools accurately, it's really easy for all of that information to simply vanish.
• 2:06
  SPEAKER_00: I know someone who accidentally deleted one of their agents that they had been working on.
• 2:12
  SPEAKER_00: They deleted all of their skills in Claude and had to figure out how to get them back.
• 2:17
  SPEAKER_00: Spoiler alert, if you're doing any of those kinds of uh things within these tools, you want to make sure you find that backup or export section because if you're not doing that on a regular basis, if you accidentally delete something, you may not be able to get it back.
• 2:33
  SPEAKER_00: But worse than accidentally undoing your own work, is those platforms can change how they're approaching the way that you do the work at any time.
• 2:43
  SPEAKER_00: The big example of this right now is how Chat GPT decided that hey, we're not going to have custom GPTs anymore, even though those custom GPTs were actually similar to Hotmail having that signature line that says get your free email in email box on hotmail.com.
• 3:03
  SPEAKER_00: Same kind of thing.
• 3:04
  SPEAKER_00: People were creating these custom GPTs that they were training and they were sharing them with their audiences, sharing them with their workplace to show people here is what you can do with AI, here's the potential that you could be realizing.
• 3:20
  SPEAKER_00: But now they don't really need those for selling them.
• 3:24
  SPEAKER_00: We're all very aware of the risks, the benefits of using AI, and so now things are different.
• 3:32
  SPEAKER_00: So if you've put all of your efforts into creating custom GPTs or gems or skills and projects in various tools, it's time to maybe take a step back and think about documenting what you're doing so that way in the event that the platform changes things up, that you're not left starting from scratch again because that would be really frustrating and you know would negate any savings of time or resources that you can attribute to your use of AI for marketing.
• 4:04
  SPEAKER_00: Alright, so with that in place, I'm gonna walk you through what is in my AI instruction system so that way you're really clear on what to put in your version of this that you create for your teams to ensure that everybody is approaching their AI use the same way each time.
• 4:24
  SPEAKER_00: And what's great about this is when you create a template for each type of asset that you regularly create, that you have templates for, you do it one time, and then somebody very quickly edits it to reflect what the name is of that new messaging document or creative brief and which template they're using, and bam, there they are.
• 4:47
  SPEAKER_00: They're ready to go ahead and do whatever it is that they need to do without reinventing the wheel and without going off-brand.
• 4:56
  SPEAKER_00: So this system is broken into three specific areas, and the first area is everything around the core identity and the goal, what it is that you're trying to do and what it is you're asking the AI to do.
• 5:10
  SPEAKER_00: And this is important because if you don't tell the AI what it is that you want it to do, what its role is, it can very frequently jump several steps ahead of where you want to go and give you something that's not what you need.
• 5:24
  SPEAKER_00: For example, if you're looking to be co-creating something with the AI and you're wanting to use it as a thought partner and somebody to bounce ideas off of, you don't want the AI tool to spit out a draft document.
• 5:39
  SPEAKER_00: Or similarly, if you're uploading your version of something, then you're gonna be asking it to maybe act as a developmental editor as opposed to take what I wrote and rewrite it for me.
• 5:50
  SPEAKER_00: Regardless of what it is that you want to do, this is the approach you take.
• 5:53
  SPEAKER_00: So that way you're very, very clear.
• 5:55
  SPEAKER_00: So the first section is where you have the persona and the role, and this is where you're telling the AI tool what it is that you're asking them to do.
• 6:09
  SPEAKER_00: Who are they in this scenario?
• 6:13
  SPEAKER_00: What is it that they're going to be doing?
• 6:15
  SPEAKER_00: So an example would be you are a meticulous copywriter, okay?
• 6:21
  SPEAKER_00: So then they know that their job is to be writing, but you could also have it be copy editor, whatever it is that you need to be doing right there.
• 6:31
  SPEAKER_00: And then you could write to reflect the persona, then you could have it say you have a deep understanding of whatever your persona is that you're writing for.
• 6:41
  SPEAKER_00: So let's say that your persona is content marketing leaders, then it would be you have a deep understanding of our content marketing leader persona, or you could say of the uh the West Coast content leadership audience, whatever it is that you've either personally you have that persona in place, or you have documentation around how that audience reacts to your content.
• 7:12
  SPEAKER_00: Whatever that is, you're going to be specifying that right up front.
• 7:15
  SPEAKER_00: So that way you're making that absolutely clear for the AI tool.
• 7:19
  SPEAKER_00: Next, you're going to either be describing what your brand's voice, tone, and style is like, which could be something as basic as your tone must be professional and direct, and you're going to use APA style.
• 7:34
  SPEAKER_00: That would be your bare bones minimum kind of direction in this area.
• 7:39
  SPEAKER_00: What I'm hoping that you'll be able to do is instead say refer to the brand voice and style guide documents that I'm uploading here, or you're hyperlinking those words, and it's pointing to a place where the tool can access them.
• 7:56
  SPEAKER_00: For example, if you are using Copilot, you can be having those documents with a link that never changes, hyperlinked in this template, and then it will point to whatever the latest version is of that of those files on SharePoint.
• 8:11
  SPEAKER_00: So again, making sure that you're telling it where to go to find out about that brand voice and about your style guide that tells it everything it needs to know about all the usage quirks that are specific to your organization.
• 8:29
  SPEAKER_00: Third, you're going to tell it what your goal is.
• 8:33
  SPEAKER_00: So, for example, you might say your main goal is to help the user draft compelling email marketing copy.
• 8:39
  SPEAKER_00: If this was for an email, um, an email that you're going to be writing.
• 8:45
  SPEAKER_00: It could also be that you're asking it to review something, you're asking it to edit, whatever the thing is that you're asking it to do, you're putting that there, and then you're telling it, what are you trying to accomplish with the activity that it's doing?
• 9:03
  SPEAKER_00: So, in the example of you're having it write an email, you are looking to create an email that drives attendance at a webinar, for example.
• 9:14
  SPEAKER_00: Maybe that's your goal.
• 9:15
  SPEAKER_00: Or your goal could be that you're trying to move somebody who put something in their cart but didn't actually check it out.
• 9:24
  SPEAKER_00: You're trying to entice them into actually hitting that button by giving them a discount.
• 9:30
  SPEAKER_00: Whatever the goal is, you put that there in the primary goal area.
• 9:34
  SPEAKER_00: And then the last part of this section is where you're going to give some expertise.
• 9:41
  SPEAKER_00: You don't necessarily have to, but if you are talking to a very specific segment that you have specific knowledge around or that you have a specific framework you want to use, you'd list it here.
• 9:55
  SPEAKER_00: So the example I like to use is you're an expert in SEO best practice and EEAT guidelines.
• 10:02
  SPEAKER_00: So if you're writing an email, you'd want to make sure that they knew all those good things, right?
• 10:07
  SPEAKER_00: Okay.
• 10:08
  SPEAKER_00: Next section is task execution and workflow.
• 10:12
  SPEAKER_00: So this is where you're actually starting to define your step-by-step process that you want the AI tool to follow when given this request.
• 10:23
  SPEAKER_00: So you're starting with the input requirements, which may be that you need the AI tool to ensure that it's that it understands the instructions, or that you are wanting to make sure that they ask questions before they get started rather than making assumptions.
• 10:47
  SPEAKER_00: Whatever the kind of thing is that it needs to have before it can operate, you're gonna want to put it in.
• 10:54
  SPEAKER_00: Frequently, it's gonna be that you need it, you need to make sure that you have uploaded a creative brief or a messaging document because it needs to understand where where the context is, where the source material is for what it's doing.
• 11:08
  SPEAKER_00: Because if it doesn't have that, it's gonna make it up, and you do not want it to be making it up based upon the bland universe of content that it's sucked into its system.
• 11:18
  SPEAKER_00: That is the opposite of what you want it to do.
• 11:21
  SPEAKER_00: So an example for that would be when the user provides type of input data, you must first action one.
• 11:28
  SPEAKER_00: So when the user provides a product description, you must first identify three key features.
• 11:34
  SPEAKER_00: If you're doing an email that needs to make sure that it connects with people in a specific way, you need to make sure that what it thinks is important is what you think is important.
• 11:48
  SPEAKER_00: So you're putting that in there so it'll make sure that you are giving it the top priorities.
• 11:53
  SPEAKER_00: Next section is the step-by-step process.
• 11:56
  SPEAKER_00: So if you need it to do something in multi-steps where it's going and it's validating one thing against another, where it's stopping and showing you something before moving forward, any of those steps, you're going to put it here by saying then you must follow these steps in order.
• 12:13
  SPEAKER_00: One, two, three, and you actually say what those are, such as draft three subject line options, write the email body using the AIDA framework, include a clear, bolded call to action, whatever those things are, you're going to be putting it right there where it is readily apparent.
• 12:34
  SPEAKER_00: Then length and scope is the final section of this.
• 12:38
  SPEAKER_00: So that way you don't end up with one of those 25-page AI outputs when you're looking for something that is a hundred words or one page or whatever the length is for the particular output that you're requesting.
• 12:54
  SPEAKER_00: So this is where you'll say the final output should be length specification.
• 12:59
  SPEAKER_00: So the final output should be no more than 300 words and divided into three distinct sections.
• 13:04
  SPEAKER_00: Like whatever you need specifically from it, you're gonna say it right there for that length.
• 13:11
  SPEAKER_00: So that way it understands if it is doing something that is a large comprehensive piece, or is it doing something that is more of a summary high level?
• 13:22
  SPEAKER_00: Because that's going to affect how it's going to process the request, how much time it's going to take, whether or not it's going to go look for external research to back things up.
• 13:33
  SPEAKER_00: All of that's going to be those variables, all can be answered by you telling it specifically how much content, how long of an output that you're looking for from it.
• 13:46
  SPEAKER_00: Now, the most important section is probably the one that I think people leave off most frequently.
• 13:52
  SPEAKER_00: Frequently, we get through what I have talked about here, and people hit okay, and they say, let's go, draft it, edit me, whatever that is.
• 14:02
  SPEAKER_00: But you need to actually make sure that you spend some time on the constraints and formatting.
• 14:07
  SPEAKER_00: So that's the third section of the AI system instruction system that I've put together.
• 14:14
  SPEAKER_00: And you start with forbidden actions.
• 14:16
  SPEAKER_00: What do you want it to never do?
• 14:19
  SPEAKER_00: So, you know, what are the actions you want it to avoid?
• 14:22
  SPEAKER_00: For me, it's I always have it say never make up statistics or use excessively casual language and slang.
• 14:31
  SPEAKER_00: I usually also say don't use emojis, especially not in bullets, no emojis.
• 14:38
  SPEAKER_00: Have I said no emojis?
• 14:40
  SPEAKER_00: Because if you don't, it's gonna use emojis because it feels that emoji is the international language of marketers everywhere and writers everywhere, and while that may be true, I don't know, but it's not true for me.
• 14:54
  SPEAKER_00: That's not how I write, it's not the kind of output I produce, so I ask it to not do that.
• 15:01
  SPEAKER_00: You may have other things that are really big deal breakers within your organization, such as if you have a brand name that has a letter in the middle of the name that's capitalized, you might say never use the brand name with without that, you know, without being capitalized in this particular way.
• 15:22
  SPEAKER_00: You might do something like that to ensure that you get the correct output that you need.
• 15:28
  SPEAKER_00: The next would be the mandatory elements, you know, such as you must always include sources with a link, a valid working link, um, whenever you cite statistics.
• 15:42
  SPEAKER_00: That's one of my favorites.
• 15:44
  SPEAKER_00: You must include a risk assessment summary in the response telling the user how you know how confident the AI is that the information is accurate and how much risk there is to your organization and to your brand if it's not, for example.
• 16:03
  SPEAKER_00: But whatever those mandatory elements that you need in order to create whatever your asset is, you're gonna put them in there.
• 16:09
  SPEAKER_00: So if you need to have a call to action in a specific format, you're putting that there.
• 16:15
  SPEAKER_00: If you need to have specific disclosure language, you're saying, hey, go look at the disclosure disclosure document and pick the correct disclosure for this piece, and then tell me which one it is so I can validate it.
• 16:32
  SPEAKER_00: You're doing that all right there.
• 16:35
  SPEAKER_00: And then last but not least, you're telling it what the output format is, which is when you actually point to that template that you created.
• 16:45
  SPEAKER_00: You're saying, hey, go look at the template for our email newsletter or look at the template for our product sheet.
• 16:52
  SPEAKER_00: And that way it actually goes and it looks at the template that you've identified that you're using to show what good looks like in your organization and to reflect all of the design elements, all of the pieces of text that a user needs to specify in order to create something, and you're telling the AI tool that that is the format that it needs to be in.
• 17:15
  SPEAKER_00: It needs to use your template, it needs to reflect the example that you're providing of a good example.
• 17:23
  SPEAKER_00: So I know this sounds like a lot, but the thing is again, once you have the instruction template for any given repeatable task that people are regularly using AI to help them with, all they have to do is make tiny edits to it because they're not being asked to change the rest of the template.
• 17:45
  SPEAKER_00: That's kind of the whole point.
• 17:46
  SPEAKER_00: They're being asked merely to update very specific, clearly indicated parts of the instructions so that way people are using they're using the same resources and providing the same context to the various AI tools that they're using.
• 18:04
  SPEAKER_00: Because when you don't do that, even if everyone is using the same general prompt for doing an activity, the AI's context is whatever it is that individual has been creating on that AI platform for themselves.
• 18:22
  SPEAKER_00: Or maybe it's for what they've been doing for work, but regardless, it's for whatever it is that they thought was good enough or didn't have the AI go further with in that tool.
• 18:34
  SPEAKER_00: That is going to be the context the AI tool has for what's good and what you're looking for, instead of having that shared context that a human would have who works within your organization and understands what good looks like, that has the ability to have good judgment, to have the good taste to know when something is right or not.
• 18:55
  SPEAKER_00: The AI does not have that.
• 18:57
  SPEAKER_00: At best, you have to find those great examples, you have to provide this context and the resources, and that's exactly why I have this template and format.
• 19:08
  SPEAKER_00: And I'm going to be making this available.
• 19:11
  SPEAKER_00: Um, it'll be in the show notes, linked to a version of this template that you can put into use in your own team.
• 19:19
  SPEAKER_00: And I hope this helps people think of a very think about it in a very concrete way so that way you can get your content foundations out of that shared folder and into use across your organization.
• 19:32
  SPEAKER_00: Alright, that's it for content chat this week.
• 19:35
  SPEAKER_00: Can't wait for the next one.
• 19:37
  SPEAKER_00: Bye everybody.