In this episode of Whiteboard Friday, Britney discusses actionable AI, how to get started, and the steps needed to make that happen.

Click on the whiteboard image above to open a high-resolution version!

Hey, MOZ fans, Britney Muller here, to talk to you about everything Actionable AI, how to get started using AI in your workflows, in your life, and really what are the steps and things you need to consider to make that happen and do it in the right way.

So, diving right in. You know, using artificial intelligence, it's really this assistive technology. Oftentimes, you're not gonna get completely from A to Z, it might just get you about 60, 70% of the way there, and that's okay, right? So kind of reframing AI as more of this assistive technology is really, really helpful.

The biggest word of warning I have for everyone, and this happens to a lot of very smart people, it happens to a lot of inside teams, is you have these over-generalized ideas or tasks for AI. You sort of have this magical orb solution that you wanna use AI for. And unfortunately, when building things out and working with AI, you need to get really, really specific. You need to boil down your ideas into bite-sized tasks. So, you know, try not to fall into that trap.

And typically what ends up happening is this happens when people are overeager to incorporate AI into their products or their systems and company. And I'm hearing a lot of SEOs, a lot of marketers, feel that burden from upper management. And there's a really neat opportunity for those of you experience that, to sort of manage up and help educate some of these stakeholders about what AI is qualified for and what it's not qualified for. And to certainly bring these tasks a bit down to Earth, in terms of making them, you know, breaking them out into steps.

So what ends up happening is people get frustrated when they try to do this, and it leads to feeling a bit defeated. And I don't want that to happen to you, 'cause there's a bunch of really great applications. I highlight a bunch of the things that Generative AI specifically is really qualified for in these blue asterisks around the board. And so to start to think about that, right? But this is an iterative experience. So again, don't get discouraged if you don't get it right on the first try. No one gets it right on the first try, I promise you.

And it's okay to think up some tasks and ideas that you wanna experiment with AI with and not end up using AI. That's okay, it's part of the process. And in fact, so many AI researchers and teams I've worked on, we do that ourselves. And so that, again, that's kind of part of the process, and don't feel bad about that. Certainly, don't stick an AI, you know, sticker onto something that you don't use AI for. But to just be mindful of that.

I like to compare this to throwing spaghetti against the wall. You know, see what sticks, see what doesn't. Some of the noodles that don't stick are things like task tracking, where you can have a heuristic model, do something like that. And all heuristic means is an, if this, then that. I know that was sort of, you know, this tool's been around forever, but just that thought process, it's sort of like a Decision Tree. If a Decision Tree could figure out your task, this is better suited for something, a different type of solution. And that's okay.

Identifying high Click-Through Rate Pages. I've heard several people wanting to use AI for that. AI isn't required to do this, right? A simple Pivot Table would get you that information quickly, and you know, not cost as much money.

Forecasting traffic, there's other really powerful forecasting models and time series models, like Facebook's Prophet, that can help you do that really, really well. And again, to sort of break down these tasks and ideas into specific parts and pieces. Let's jump over here for a second.

So in addition to this, let's consider what is AI even qualified for, right? What is it inherently good at, and what is it inherently bad at? There is some graphics maybe we could pull into the post that are on my Large Language Model Guide on Data Science 101, where I try to explain what those are, so that you don't fall into the trap of trying to use this technology in ways that it's really not created for.

And again, other models that do that task. So that forecasting example of Prophet, Jess Peck has an incredible Prophet Tutorial on women in SEO. Highly recommend you check that out. And that's kind of a great intro to these other more deterministic models that do things really, really accurately. The AI we're sort of typically talking about today is mostly generative. You're generating entirely new things, which, again, can be powerful for Summarization, Titles and Meta Descriptions, Code Assistance, creating new quizzes, but they're not 100% factually based.

So to explore some other things. To also consider the biases that are baked into these models. We see companies like PGA and their social accounts on Instagram trying to incorporate AI technology in really funny and clever ways, where they extend player's headshots. And at first, this is a really interesting application. People are getting engaged with it. Until you get to the last two players, where they're players of color, and it breaks the background, it breaks the player background, and it positions them in what looks like a dump, or working on some sort of woodworking object, right? That's not okay. This is actually reflective of the biases and the issues that come inherent with these models as well. So again, just knowing how to navigate those things and knowing where you'll bump up against these limitations is so, so important.

And to start sparking some ideas. A fun way to do that, I love to encourage people to kind of leave a window open with ChatGPT or Gemini, just to have a sort of a visual prompt when you're working on things. I still do this, and I still sort of challenge myself to come up with different ways to sort of use it to help accelerate something I'm doing. And consider what takes What are some of your most repeated tasks that maybe you don't care for as much as other things, right? I've heard people talk about going through their leads, through their website, and filtering out which is spam and which are actually qualified leads. These models are really good at starting to identify some of those things.

Same with meeting notes, there's tons of tools like Otter, and I believe Firefly is another one. But that can sit in on meetings, transcribe what's being said, and automatically generate meeting notes and kind of next actionable steps, which is great, right? That's a huge time saver. And it's really allowing us to be more human, right? And that's kind of at the center of all this, is that human-centered element. And to consider with this, what unique value do you add to your job, right? What unique insights and opinions and experience do you have that you can sort of double down on, and enhance in what you're working on, right? Because artificial intelligence has no, has no experiences, it has no ground truth in the world that we live in.

This is why the human-centered AI approach is so, so critical and will be so important moving forward. To consider what's the purpose of this application? Who is this serving, right? Who will be impacted if this goes wrong, right? If we get this thing wrong, what could the potential outcomes be? Who's gonna be in the loop, right? These aren't set it and forget it, applications. Someone should always be in the loop and identifying, was this quiz generated, is it correct? Were there any strange errors that came out of it? And again, this doesn't get you all the way there, it's just getting you 70%, 75%. And then to really, really take a step back and consider what are some possible unintended consequences of this? And so when I am talking to new people about AI that are interested in incorporating this, what I've started to come up with, with my Generative AI Course, is this sort of beginner AI stack or ladder, and here's how to think about this. So it's really sort of easy to more complex.

And what I always encourage beginners to experiment with is get a Proof of Concept. Just see if these different tools are capable of doing the thing that you want them to do. So for example, go straight to ChatGPT, go straight to Gemini, go straight to the source, get that Proof of Concept, and then we can start thinking about these higher-level integrations and tools. MidJourney, is slightly more technical, just because you have to be on Discord. But what gets really exciting is things like NotebookLM. This is a free tool by Google. You can upload a source of information or data, and you can query against that. So recently, I uploaded my "Large Language Model Guide," and I created a glossary like that, right? (finger snapping) I created a quiz; you can create a study buddy for students. It's really, really powerful.

And then up from there are Platform Integrations. These are things like GPT for Sheets Plugin, so, so powerful. A student recently wanted to take product descriptions and come up with title tag, meta description, and draft up social posts for Instagram, Twitter, and LinkedIn. And so we thought as a class, "How can we best do this?" And GPT for Sheets has a formula that auto-splices it into columns. And so with just one formula that was referencing the product description, whew, we're able to automatically generate rough drafts for all of these things. Again, there should always be a human in the loop, but how exciting to start to incorporate this in ways that really help you focus on higher-level things and help you do more strategic work. And so again, always, always break your task down into specific steps and figure out if GenAI or AI is qualified for that. And to be where the user is, right? So if it's that Lead Generation Email Filtering, come up with a solution like a model through Zapier for Gmail that can auto-flag what's a potential qualified-lead versus spam.

So hopefully, this gives you more of a framework to work from in terms of how to incorporate this technology into your work, into your life. And if you're interested in learning more about this AI ladder and making, you know, more of an actionable end-to-end solution, please feel free to check out my, Maven Course. I'm helping people do that on the Fundamentals of Generative AI.

And thank you so much for watching this, "Whiteboard Friday." This has been so fun. If you have any comments or questions, please feel free to tag me online, whether that's X or LinkedIn and Instagram. Let me know if you have any thoughts and I look forward to seeing you all soon. Thanks.