Paving the Way: Strategies for Integrating AI in Marketing

Many marketers today recognize the potential of AI but aren’t sure how to start integrating it into their strategies.

The challenge can be daunting—how do you effectively introduce AI into your existing workflows, choose the right tools, and ensure your team is equipped to leverage its capabilities? Moreover, with myriad options available, understanding where to begin and how to navigate the potential pitfalls can seem overwhelming.

Today, we’re tackling these questions with the help of Christopher S. Penn, a renowned expert in AI and data-driven marketing. Christopher brings invaluable insights from his extensive experience with AI implementations in various business contexts. With his deep knowledge in both the technical and strategic aspects of AI, he’ll share practical strategies and real-world examples to help you pave the way for AI integration in your marketing efforts, making the process more manageable and effective.

Christopher will walk us through the crucial steps to get started, how to select the right AI tools for your needs, and the importance of team training and upskilling. We’ll also delve into common challenges and how to measure the success of your AI initiatives. By the end of this episode, you’ll have a clearer roadmap for integrating AI into your marketing strategy, ensuring that your business is not only prepared for the future but leading the charge in innovation.

AI in Marketing: Unpacked host Mike Allton asked Christopher S. Penn about:

✨ Starting with AI: Learn the initial steps businesses should take when integrating AI into their marketing strategies.

✨ Choosing the Right Tools: Understand how to select AI tools that fit your specific marketing needs.

✨ Training and Measuring Success: Discover the importance of team training and how to measure the success of AI initiatives.

Learn more about Christopher S. Penn

Resources & Brands mentioned in this episode

Full Transcript

(lightly edited)

Paving the Way: Strategies for Integrating AI in Marketing

[00:00:00] Christopher S. Penn: So we exported that data out of there, combined it. with a couple of the client’s other proprietary data sources and help them create ideal customer profiles for every segment of their market based on what the audience likes, like what their interests are like movies and shopping and TV, certain types of TV shows, certain types of music, et cetera.

And made this available to them so that their marketing teams can now effectively do synthetic focus groups where they can talk to a persona and get real answers about like, what should we change in this email? That’s going to go out to X million people. What can we change in this email that will get higher click through rates that will get more interest in things.

And so we rolled this out. And the part that astonished the client was that they were going to do this a different way. It was going to be a 12 month project. And we’re like, pretty sure we can help you out there. Um, so we spun up a new scope of work and work with them and the initial customer profiles were done in 10 days.

[00:00:58] Mike Allton: Welcome to AI in Marketing: Unpacked, where we simplify AI for impactful marketing. I’m your host, Mike Alden here to guide you through the world of artificial intelligence and its transformative impact on marketing strategies. Each episode, we’ll break down AI concepts into manageable insights and explore practical applications that can supercharge your marketing efforts.

Whether you’re an experienced marketer just starting to explore the potential of AI, this podcast will equip you with the knowledge and tools you need to succeed. So tune in and let’s unlock the power of AI together.

Greetings program. Welcome back to AI in Marketing: Unpacked where I selfishly use this time to pick the brains of experts at keeping up with and integrating or layering artificial intelligence into social media, content, advertising, search, and other areas, digital marketing. Oh, and you get to learn to subscribe to learn how to prepare yourself and your brand for this AI revolution and come out ahead.

Now, many marketers today recognize the potential of AI. But they’re not sure how to start integrating it into their strategies. Challenge can be daunting. How do you effectively introduce AI into your existing workflows? Choose the right tools and ensure your team is equipped to leverage it, all of its capabilities.

Moreover, with myriad options available, understanding where to begin now to navigate the pitfalls can seem daunting. Today we’re tackling those questions with the help of Christopher Penn, a renowned expert in AI and data driven marketing. Chris brings invaluable insights from his extensive experience with AI implementations in various business contexts.

With his deep knowledge in both the technical and strategic aspects of AI, he’ll share practical strategies and real world examples to help you pave the way for AI integration in your marketing efforts. Making the process more manageable and effective. Chris is going to walk us through the crucial steps to get started, how to select the right AI tools for your needs and the importance of team training and upskilling.

We’ll also delve into common challenges and how to measure the success of your AI initiatives. By the end of this episode, you’ll have a clear roadmap for integrating AI into your marketing strategy, ensuring that your business is not only preparing for the future, but leading the charge in innovation.

Hey, Chris, welcome to the show. Thank you for having me. So excited to jump into this with you. I know we’re going to cover a lot of ground today, but to start, can you tell us about your journey into AI and what led you to focus on this intersection of technologies? Sure. So

[00:03:20] Christopher S. Penn: I think before we do that, we should even talk about what AI is because there’s three major branches of it.

There’s regression, there’s classification, there’s generation. Now we like to use the acronym FOG, find, organize, and generate. The first to find and organize that what we call classical AI. So this is stuff that is as old as like, goes back to the 1950s. We’re talking the marketers would know as things like uplift modeling, marketing mix modeling, attribution, that’s that sort of regression based generative a classical AI, organized classification is, hey, I’ve got all this data, I’ve got, you know, a gazillion and a half social media posts, help me categorize, put them in topics, bucket them and things like that.

So organizing your data. And again, marketers have seen that even something simple as like a word cloud, like that is That is classification. The third branch is generation, which uses the technologies of. regression and classification to make stuff to have AI make things. So in terms of my journey, I started in, I started marketing technology in the early 2000s, at a financial services startup.

And then in 2005, Google bought this company called urchin, which then they rebranded and started giving away as a as a service called Google Analytics. In 2012. I left the company I was at to go into PR firm because they were having some really challenging measurement problems. Like, how do you measure the impact of PR when there’s nothing to click on?

And part of that was digging going from straight statistics into data science and machine learning classical AI to resolve, okay, how do we solve this problem mathematically, and that was my first introduction to what we now call AI. So 2012 2013, that’s when I started doing a lot of coding and are a lot of statistical stuff, a lot of data science stuff.

And then, of course, in 2017, my account director at the time and I decided we were going to we we saw where classical AI was going like there’s legs to this thing. And the PR from what was going in a different direction. And so we parted ways and we’ve, we decided to switch things up. And so Katie Katie Robert and I co founded trust insights and we launched in 2018 and our focus has been on.

analytics, data science and AI. That’s where we are now. And of course, you know, in, in November 22 open AI pivots, what would then was it’s a GPT three model into that was in the playground into production with a model called instruct GPT. And we all the general population had learned about through a service called chat GPT.

And then suddenly everyone’s an AI expert

[00:06:05] Mike Allton: suddenly overnight, right? But I appreciate you sharing that because we’ve known each other for a long time. I think I met you in person 20, 16, 2017, something like that. 2015 content marketing role. And I, I, even to that point, I did not realize how invested you were in, in the concepts of AI.

So long ago, because because to your point, a lot of us and myself included have pivoted into AI extremely recently, like within the matter of days for some of us. This has been your playground for a while.

[00:06:39] Christopher S. Penn: Yes, this has been the thing about machine learning and artificial intelligence is that when you when you take away all the hype.

All of these things are just prediction engines based on probability. And if you understand the architecture of how these systems work, you can make them work better. I mean, that’s, that’s fundamentally, when you look, for example, at the inside workings of a transformer based model, like the ones that power chat GPT, and you read the academic papers on stuff, you start learning about the architectural features under the hood.

And the analogy I likened to is a blender. Most people know what a blender is, and they know what you should put in it, but you shouldn’t put in it like if you’re making steak, wrong tool. But that’s about it. A certain subset of people know that a blender is actually an electromagnet, you know, there’s electricity, there’s wires, copper coils, that drives an axle that you know, blends things, and they can take that apart, they can understand, okay, well, the blender is not working.

Well, here’s why this part of the drivetrain is broken. When we think about AI. and maximizing performance. And to your point, integrating it into marketing into enterprises, you’ve got to go from I know what to do with a blender to, you know, can I fix the blender if it breaks? How should blenders be rolled out across our company?

who should be allowed access to the blender, right? Don’t don’t don’t give access to person who’s got to stick their hand in there. And knowing when is the blender the right tool. So there’s a lot of that that goes into these things. And then how do you make the blender part of the workflow? How do you connect other tools to the blender?

how do you take the outputs of the blender and use them? For example, if you’re making steak and you want to tenderize a steak, take a blender, take some pineapple, blend that pineapple up, put it on the steak and let it sit for an hour and you will have really tenderized steak because the bromelain inside pineapple actually digests some meat proteins.

Pineapple is the only food that when you eat it, it eats you

[00:08:42] Mike Allton: Okay, this is an analogy that can be applied in a lot of different ways, but I appreciate the reminder that people are different stages of understanding this technology, because it’s something I have to constantly remind myself when I’m talking to my audiences. I studied computer science In college, and I used to be an it sale.

So I’m watching, you know, Jensen’s keynote address for the NVIDIA developer conference, and I was understanding virtually everything he was talking about with, with chips and semiconductors, all those kinds of things. But not everybody else understands that everybody else needs to. So those of you listening, you don’t necessarily need to understand all of these things, but Chris’s analogy is.

It’s perfect. You need to understand what the blender is, how it works, when to use it. So let’s get into that, right? What are some of those first steps that a business should take when they’re kind of considering integrating AI specifically into their marketing strategy?

[00:09:34] Christopher S. Penn: So unsurprisingly, I’m going to go ahead and share my screen here.

We have a whole framework for this. So my company trust instance, we do this as literally as consulting. So the, the, the base of the framework is purpose. Why are you doing it? Right? What is the business purpose for doing something? Because a lot of the times, and this is especially any hype cycle, there’s a lot of people who like, we’ve got to use AI.

You know, why, right? What are you trying to do? What’s the business outcome you’re trying to do? And then at the top of this is what. are you going to use to measure that AI was the right choice, right? If you don’t have those two bookends, it’s like a sandwich, like it’s like a BLT. But you don’t have the bread.

So like, okay, I’ve got a salad. This sucks. Even the salad has bacon in it. It’s not what you wanted. And so you’ve got to have those first few parts down. And anytime you’re thinking about AI applications, you’ve got to have your purpose, you’ve got to have a system of measurement that tells you that you did the thing.

Generally speaking with AI. any form of AI, you’re looking for one of three key outcomes, save time, save money, or make money, right? And that’s, that’s it. Like if you can’t connect the project, the idea of the use case to save time, save money, make money, it’s do a it should be because you’re just going to waste a whole bunch of time.

Now if you’re doing it, if your job is R& D, totally different than it’s like, what can we learn? But even there, what can we learn still should have some line of sight to save time, save money, make money. So that’s the first part. Second part is there’s there’s sort of a hierarchy of the use cases for AI, which is the process column.

There’s at the very baseline, individual prompting usage, hey, we’re using chat GPT, right? That’s pretty much like 98 99 percent of people are there today, and that’s fine. individual employees, consumer AI interfaces, like chat GPT, that’s where you got to start. Because you need people to buy into it, you need executives to see, okay, it’s not going to end my company immediately, right.

And building that that sort of comfort, the second level systematized prompt management, now we’re starting to talk about scaling things, right. So instead of you using island of your own, you and your team have some sort of knowledge management software, some kind of knowledge management store, there’s, there’s some project management around that.

So hey, we’re going to use this to generate content, we’re going to use this to process customer complaints, we’re going to use this to analyze strategy, whatever the thing is, you’re now starting to have that systematization of your prompting, maybe some of your data sets a little bit. Your third level is now starting to simplify the UI and standardize.

Now we’re talking about building interfaces around these things. So for example, custom GPT is in in chat GPT, where you can build a purpose built tool, you simplify what people have to do in order to interact with an application. Your fourth level advancement is we’re building API’s and full apps, Google Vertex, for example, where you’re going to build a chat bot that internally is in your slack.

And like, for example, the one thing we do, we just do this for customer, we built an ideal customer profile, synthesized all the data built ideal customer profile, and then put it into a vertex chatbot in slack. So now when marketing goes, Ah, you know, I wonder what the customer think of this, you talk to, I think we called ours, Emily Thompson, you just chat at her in Slack and say, Emily, what do you think of this email that’s about to go out?

You know, how would you feel as a customer? And she goes, ah, you know, it doesn’t really appeal to me. It’s it’s, you know, and they say, Okay, well, what would appeal to you? Well, maybe she said it like this. And for people that’s now now you’re taking away generative AI as a thing you have to do and making it available somebody as a service, right?

This is a this you’re building a service like okay, I can talk to this person. After that you start getting into data engineering. Like what data do you have as a company? that you can leverage in these tools to make better personas to make better strategies, etc. That’s second to the highest level is fine tuning custom models.

And you found that one thing that is so mission critical that you’re going to build your own AI around it. And that the height of your evolution AI is everywhere. And it’s like electricity, you don’t even think about it anymore. It’s just part and parcel of you as a company. So the framework for how do we think about rolling out AI integrating into our marketing looks like this.

And you can see in the people column, it’s not just marketing. If you want this to succeed, you’re talking about integration, across multiple departments. If you look at the platform that you’re talking about integration across multiple tools, you cannot no marketing team, no marketing department can fully see the benefits of generative AI without collaboration.

[00:14:28] Mike Allton: And just as a quick note to those of you listening, we’ll have the links to everything that Chris was talking about. And these, these slides in the show notes, if you’re not if you’re listening on your phone, we’ll have that for you. But A couple of things that you said that really stood out to me first.

I love the use case of having that slack integration. I’ve been imagining something a little bit similar for our sales team at Agorapulse. We have a tremendous wealth of resources, blog posts, podcasts, video snippets, and so on that speak to specific pan points that our target audiences. Are going through and we want to equip our sales team with those resources, but once you pass like 10 links, it becomes a little complex and almost impossible for an individual age.

Remember everything that we’ve ever created on every platform that might speak to any potential pain point. And so I’m thinking, well, this is perfect for AI. And I don’t know if that that slack integration Or that Vertex, how did, what was it? So

[00:15:24] Christopher S. Penn: it’s Google’s Vertex AI studio is the environment you build that in.

[00:15:29] Mike Allton: Yeah. And you think that would probably be a good platform for that kind of a solution as well.

[00:15:34] Christopher S. Penn: Maybe. It depends on your sales team. It depends on your sales team. CRM and infrastructure because obviously different CRMs have different capabilities, right? HubSpot has chat spot. Salesforce has Einstein, everybody and their cousins, you know, to some degree appropriately is trying to figure out how do you integrate generative AI and natural language prompting in complex software because That’s a heck of a lot easier to, you know, when we show this to, you know, sales executives we say, you know, which would you prefer to do, you know, 28 clicks to get to your dashboard for your pipeline or go to chat spot and say, what’s my sales pipeline look like today?

And it just says, here’s, here’s sales pipeline. That’s so much more natural for people.

[00:16:13] Mike Allton: That’s a really great point. Insight. We are using both HubSpot and Salesforce, but I’m pretty sure the sales team is focused on the Salesforce side. So we’ll have to look into using Einstein, but that leads me into my next question with all these tools that are not only being created as AI powered tools, but like, In this case, Salesforce, they’re integrating AI into every aspect of the tool.

How do business go about approaching and selecting the right AI tools or platforms for their specific needs?

[00:16:44] Christopher S. Penn: Unsurprisingly, this, this framework is your blueprint for that, because if you know where you are. And you know where you want to go, you can then start doing formal requirements gathering to ask, Okay, what, what do we want to be able to do?

Right? So the one of the biggest challenges with with artificial intelligence, particularly generative AI and language models is that language, name a single function in your company that does not use language, right? There isn’t. it’s it’s it is literally how we communicate as human beings. And as a result, anything that is language is an opportunity to use a language model to do.

So the use cases are literally infinite, there’s there’s no shortage of them. In terms of the technology, the technology is what you select last. You’ve gotta have, who are the people involved? What are the processes that are, that are in place right now? There is, lemme go find it real quick. The over on our website we have a, a framework that we call trips.

And it stands for it’s an a free download you can get with, there’s not even a form to fill out because. We’re trying to inter share as much basic stuff about AI as we can. So the trips framework is this five categories that you score any given AI task or any AI given task as to whether should be AI or not.

Number one time, how much time does the task consume updating Salesforce? takes a lot of time, right? That’s and that’s a big one to how repetitive is the task. salespeople have to update their CRM, right? As something you do, you’re supposed to do it every day. Some most, you know, SDRs and stuff will do it like once a week, maybe, right?

But it’s highly repetitive. Three, how important is the task? This is one’s a bit of a mental twist. The less important the task is, the better because the more important it is, the more it’s going to likely need human review, right? You don’t need some a bunch of tasks, you just don’t want to hand off to a machine just walk away from number four, how pleasant is the task who loves updating Salesforce?

No one, right? Less pleasant task is the more candidates for AI. And fifth, do you have examples of how the task is supposed to be done? How many examples you have, the more examples you have, the better, that’s sufficient data. So this trips framework is a way to do your look at your processes, score your processes, and then Once you’ve done that, you basically have the requirements you can say, Okay, well, now we need to find a vendor that will do these these sets of tasks.

And here’s how the task works. And here’s how we do it right now vendor, show me how you will your tool maps to this. And if you want to get fancy, you build your requirements document, right? Any brd anyone who’s ever done a brd, you know, business requirements document, you want to flush that out. Here’s what we here’s like, what you must have nice to have, don’t want, etc.

And that goes goes in there, then you get clever, you go to a tool like Google’s Gemini or anthropocloud three and say, here’s my business requirements, build me a scoring rubric. for this of how to assess it. And then you issue your RFP. And then you get the vendor RFP responses and you say score each RFP according to my rubric for my requirements.

And let’s build the shortlist of which vendors satisfy the requirements and we’ll, we’ll dramatically speed up the joy of RFPs. And select a vendor that, that matches the requirements.

[00:20:21] Mike Allton: I love that idea. We’re talking about a new tool selection here at Agorapulse and not AI related the tool itself, but the process that we’re going to go through, you just outlined, I think it’s going to save us a lot of time and you reminded me of this fantastic beam that’s going around where the woman says, I don’t need AI.

To create images and write for me. I need a I to do my laundry, do my dishes, speaking to the things that are unpleasant that we don’t want to do that. We’d love for systems like a I to take over. But you also have me wondering when you’re talking to businesses about how they’re integrating a I, this seems like something that not only should people be talking to about, but There should be someone who’s actually making decisions on behalf of the company, on behalf of departments.

Are you seeing clarity in who those individuals should be within organizations,

[00:21:16] Christopher S. Penn: who they should be and who they actually are two different stories, who they should be is your AI task force, right? So you would have an AI task force of stakeholders. from three different levels in your organization, you have your executives, you have your management, and you have your staff and you have so that’s through sort of the horizontal if you can broadly imagine that in your organization, and then you have your verticals, HR, finance, marketing, sales.

Ideally, your task force has one person from each level in each vertical. So that as a collaborative team to say like this is what we want to do. with this technology. Everyone do your trips framework, everyone outline all your tasks, gather up all of our information and figure out what are the priorities of the organization, what things what things are just time sucks, right?

I worked at the PR firm, there was this one position account coordinator at the firm, that person’s job was to copy and paste Google search results into a spreadsheet for clients 40 hours a week. That is a massive time suck. That is a waste of human potential that that doesn’t even need AI like that’s just basic API automation, but they did not have the capacity to do that.

With generative AI, you can write the code, not be a coder, write the code to do that, run the code and then remove that mind numbing unpleasant task and get that human skilled up on like client relations. Who among all of us would not want more attention and better service from our vendors, right? To have someone who will answer the phone when we call?

Well, if you free up people from doing Boring shit like that. And it’ll copy pasting all day. When you pick up the phone, they can, you know, someone can answer.

[00:23:02] Mike Allton: Absolutely. And so I think we touched right there on one of the major challenges facing businesses is, you know, who do they tap? How do they organize that?

They’re struggling with that. Individuals within the organization are making decisions for themselves. Without any kind of oversight or communication or education on the basics of AI tools, let alone anything that’s a little more complex in terms of how it’s going to benefit the business. But what are the challenges?

Have you, are you seeing businesses face when it comes to implementing AI?

[00:23:31] Christopher S. Penn: There’s a bunch. So the, the challenges again, go back to people process and platform platform is almost never the challenge. Well, actually, that’s not true. A lot of companies have, they don’t have a good understanding of what the foundation tools like chat GPT or Claude or Google Gemini, what they’re capable of.

So they end up buying a whole bunch of software from a whole bunch of vendors that promise them magic. And the reality is 80 percent of the tasks that you can do with a for using generative AI you can do with a foundation tool, right? You don’t need to buy a vendor’s tool. So you certainly don’t need to buy a vendor’s tool to write blog posts, right?

That is that is a complete waste of money. So that’s a big part is understanding what the technology is capable of on the platform side. But the bigger obstacles are people in process. So with process, what data do you have? Where is it? What condition is it in? Can you use it? And there’s there’s not a whole companies don’t really have a good handle on they have not had a handle on that ever.

One of the things my co founders Katie says often is that new technology doesn’t solve old problems, particularly when the problems are people, you know, take something like a sales executives, right sales executives. don’t like updating Salesforce, this is not a new problem. And no amount of technology is going to solve it, you can have things that are easier, right?

You can say, Okay, hey, you know what, instead of updating Salesforce, just leave me a voicemail. And we’ll have AI transcribe it move that into Salesforce. But at the same time, you still have to get that person to want to update Salesforce. And, you know, there’s there’s a variety of incentives and things, but that’s not a technology problem.

That’s not an AI problem. That is a you hired a recalcitrant salesperson problem. And that’s that is the challenge across the board when it comes to AI integration and AI in general is some people assume it’s magic. It’s not magic. It’s a prediction engine. Some people assume that it’s going to take their job and about 40 percent of people are correct.

And Some people assume that the technology is just all hype, and that’s also incorrect. So there’s a wide spectrum of less correct beliefs about the technology. And the critical thing that companies should be doing is understanding what the technology is and is not capable of identifying people within the organization who are already using it.

hint, it’s three quarters of your organization. At least according to the Microsoft 2024 work trend index reports, three quarters of employees are using it with or without permission, which means that your corporate data is leaking out to other places. Data security and data privacy are obviously considerations.

So you need to be thinking about governance, how you’re going to govern not just use AI, but use of your data with AI. And then there’s the legal side of things. What can you and can you not use AI for? For example, it is pretty look almost across the globe. In most countries, under the law, machine generated content cannot be copyrighted, right?

Because machine made in the US 2018 case, Nero versus Slater, a chimpanzee took a selfie, this went to court, the photographer tried to copyright this went to court, and the court ruled the chip did the work, not the human. And chips can’t hold copyright, because they can’t defend it in court because they’re chips.

And therefore, this pictures in the public domain, it cannot be copyrighted. This has been extended to human to AI, you know, chat CPT will not show up in court to defend itself. And so machine generated content can’t be copyrighted. If you are a company, say that you’re a marketing agency, and in your contracts, it says you all work you provide is work for hire and you assign copyrights to the client.

you cannot assign a copyright to a document that does not have copyright, you you are technically in breach of contract. So you need to have a sit down with your legal team, and you have to sit down with your clients, you have to sit down and explain like, here’s what’s inbounds and out of bounds. it is probably a perfectly okay use case for your monthly summary of your marketing activities to be AI generated because you hand that to the client.

They’re not going to put it on their website, right? It’s an internal document, totally fine use case. You hand an ebook that you wrote with AI and the client puts it up on their website. It’s very obviously written by machine, the competitor steals that the client tries to sue and the competitor says this is AI generated, you can’t sue me because there’s no copyright.

Now you you have put your client in a hot spot. So all these are considerations that challenge companies when they’re thinking about integrating AI into their workflows that you need to have a good task force that is familiar with the issues. So that they know what to do about it.

[00:28:16] Mike Allton: Fantastic example of why that is so important.

And it’s funny, you said AI is not magic. Arthur C. Clark might argue that it’s sufficiently advanced technology to appear as magic to many of us. So give us an example of a use of AI in marketing that perhaps you’ve been involved with that that’s been very successful and, and honestly, it might even appear magical to some.

[00:28:43] Christopher S. Penn: So recently at one of our clients they, they needed to have a way for their marketing team and their market research team to be able to efficiently understand the needs of their customers. is a big, big B2C company. And I said, we, we have data here somewhere we don’t know where we don’t know what we have.

And I said, Great, do you have Google Analytics? Like, yes, we do somewhere. So after a little bit of poking around, we were able to get access to Google Analytics. And what a lot of people don’t know is that inside Google Analytics, thanks to Google’s ad systems, you can get things like demographic data about your audiences, get what they’re interested in, their interests and affinities and stuff.

So we exported that data out of there, combined it with a couple of the client’s other proprietary data sources, and help them create ideal customer profiles for every segment of their market based on what the audience is. what their interests are like movies and shopping and TV, certain types of TV shows, certain types of music, etc.

And made this available to them so that they can then their marketing teams can now effectively do synthetic focus groups where they can talk to a persona that is assembled by generative AI from their proprietary data plus Google’s data. And get real answers about like, what should we change in this email?

That’s going to go out to X million people. What can we change in this email that will get. higher click through rates that will get more interest in things. And so we rolled this out. And the part that astonished the client was that this was, they were going to do this a different way. It was going to be a 12 month project.

And we’re like, pretty sure we can help you out there. So we spun up a new scope of work and work with them. And the initial customer profiles were done in 10 days.

[00:30:36] Mike Allton: Wow.

Any, anyone who has waited and waited and waited for the department, the company to create those kinds of ICPs can relate to the stunned factor of that particular project.

That’s amazing. It was a lot of fun, really cool. Yeah, I bet. And I’m sure they were, they were pleased. And I’ve seen, you know, to your point earlier, I’ve seen lots of tools and startups claim to be able to create ICPS for companies. They’re talking about bringing in data from across the web, but it’s not personalized to that company.

They’re not looking at that particular company’s analytics and analyzing it the way that you just described. It’s just it. general, if I’m in this industry, and I’m this kind of business, who am I talking to? And that’s nice. But I think it’s so much more powerful if they’re using the actual data that’s available to them.

[00:31:29] Christopher S. Penn: And here’s the thing. This is something I end my keynotes with. There’s two things that will set you apart in in success for generative, all forms of AI, but generative AI especially. Number one, is the quality and quantity of your ideas. These tools are skill levelers, right? They, they can help. There’s a study from BCG and Wharton last summer, that showed that when given AI, the in a to 700 consultants, the people in the AI test group who are in the low 50 percent of the underperformers, when they used generative AI, they became they scored higher on 14 separate tasks than the top 50 percent high performance or turn low performance at a high performance in five hours, right?

That is a huge performance jump. And so when we’re looking at these, you know, successful integrations and things, it really is just about getting people to, to understand what’s possible. So their skill levels, their skill levels, which means that one of the big differentiators is the quality and quantity of your ideas, whoever has the most best ideas.

will win because generative AI will do the actual work. The second key differentiating factor that will make or break your AI efforts is particularly in particularly in marketing is whoever has the most best data. If you go into a tool like Gemini or chat GPT, and you say let’s, you know, let’s write a blog post, we’ll do a pedestrian exam, let’s write a blog post and give it an outline.

And it does the thing like, okay, great, I wrote the outline. That was cool. If, on the other hand, you had, say 28 academic papers on the topic, you could bring those in, merge them together, synthesize an outline based on that specific data, and create something that is so unique and compelling. I did this last weekend for fun.

Of course, I was. Well, I was really I saw a post on LinkedIn that really pissed me off. It was this LinkedIn guru, learn the seven secret hacks that should be illegal on LinkedIn. I got this bullshit. And so I went in to LinkedIn’s engineering blog, I extracted 70 blog posts from their engineers that they openly talk about how the LinkedIn algorithm works like the different technologies.

I went to archive and I pulled 28 academic papers that LinkedIn engineers had submitted to different conferences about how they use AI LinkedIn, and how their systems work. I pulled several interviews with Tim Yorker, their head of engineering, that he’s done on various podcasts like this week in marketing and AI.

This week in machine learning AI, I grabbed all this information shoved into Gemini was seven ish hundred thousand words, give or take. And I said, from all this synthesize how LinkedIn’s algorithm actually works. And it did a great job. I said, Okay, show me all the technologies first. I said, No, I post something on LinkedIn, walk me through every single step of what happens to that post on LinkedIn, what system interacts with how it scores, how it rates and stuff like that.

And it did that. So good, great. Now you’re going to give me a basic LinkedIn marketing strategy to increase the visibility of my content. in the LinkedIn news feed, and it spit out a nice outline. I said, Great. Now, I want you to give me an intermediate one. it’s all right in the basic show me the intermediate stuff and gave me a new series of recommendations.

It’s a great now let’s assume I’m a really advanced user. Don’t show me anything that you’ve shown me in basic and intermediate show me just the advanced stuff. And it gave me some very technical things to do with LinkedIn. And I suddenly had a 20 page book on how LinkedIn works. We’re actually giving you’re giving this way it’s on it’s on the trust instance website.

If you want to grab a copy of it’s totally free. Then Because I love poking the bear. I went to YouTube and I grabbed 20 different videos from the so called LinkedIn gurus, you know, the secrets you need to know, I grabbed all the closed captions, all those videos, I put them into Gemini two. And I said, Now, compare and contrast the advice that the gurus are giving with what LinkedIn engineering actually said.

And there’s a long document that I sent to some friends who work at LinkedIn. And Anytime you see that, you know, one of these gurus spouting off, feel free to slap them with, and here’s how you’re wrong about how it actually works. Because it turns out there is no such thing as the LinkedIn algorithm.

There’s 13 separate independent AI systems that all work together to, to decide what is newsfeed. So your so called hack really you know, any results you’re getting are mostly coincidence. And oh, by the way, just like Facebook, LinkedIn recompiles its algorithm on like an hourly basis across its 13 petabyte servers that are keep the entire network graph in memory and synchronized across the planet.

Like the whole system is astonishing and how advanced it is. how fast it is. And the, you know, these, these, these so called gurus understand none of it. And that’s an example how people should be thinking about using these tools like that is an advantage. If you know where the data is, And you can have these tools synthesize it and find connections in them that you can’t find.

Because your brain my brain can’t keep 700, 000 words of information in memory all at once. These tools can do that. It’s incredible. And it’s a it’s a such a powerful advantage for you as a company. So long story short, the quality and quantity of your ideas. We’ll make you win with AI and the quality and quantity of your data that you bring to the party will make you win with AI.

[00:37:00] Mike Allton: Yeah. And it’s fantastic how these tools are getting stronger and better and more powerful every single day. We, again, you know, throw back to that, that Nvidia keynote, we saw that, you know, that they’re, They’re processing power that, that, you know, we’re going to have at our fingertips, literally within a matter of months, as these new chips get rolled out, it’s going to be phenomenal.

And I appreciate you sharing a couple of use cases and stories, because that’s one of my purposes with this show, those of you listening. Stick around because you’re going to hear so many of those kinds of use cases in the future episodes. And I think that’s how we’re all going to expand our understanding of what’s out there, what’s possible.

Folks, we’re talking with Chris Penn about integrating AI into your marketing. And there’s one question coming up in particular that you’re not going to want to miss. But first, let me share with you the tool that I’ve been using to help me integrate AI into all of my most important tasks. This episode of AI and Marketing Unpacked is brought to you by Magi, your gateway to making generative AI incredibly simple and accessible.

Wondering how to seamlessly integrate AI into your marketing strategy without getting bogged down by complexities? That’s exactly where Magi shines. It provides user friendly AI solutions that empower marketers just like you to innovate and elevate your campaigns without needing a degree in science.

Imagine having the power to generate creative content. Insightful marketing data analysis. Or even personalized customer communications, all at the touch of a button. Magi isn’t just about providing tools. It’s about transforming your approach to marketing with AI that’s tailor made to be straightforward and effective.

So, whether you’re looking to boost your content creation process or want deeper insights into your marketing performance, Magi makes it all possible with a few clicks. No fuss, no hassle, no hassle. Just results ready to simplify your AI journey. Visit Magi today to learn how their solutions can revolutionize the way you engage with your audience.

Don’t just market market smarter with Magi tap the link in the show notes. So Chris, how can marketers ensure that their AI initiatives align with their overall marketing goals and strategy?

[00:39:11] Christopher S. Penn: It comes back down to the, that 5P framework that Katie invented. If you haven’t figured out what the purpose is.

that’s aligned with those goals, then yeah, your AI initiatives going to go nowhere. Because at some point, someone’s gonna say, what are we spending all this time and money on? And what do you have to show for it? If you start out by saying we’re going to take the role of account coordinator and free up 40 hours a week.

And again, free up that person. Then when you do that with AI, people like, wow, you did it. Good job, right? So you have to be clear on the purpose and performance of any project, not just a up any project up front, so that you know what you’re trying to get. And the easiest way to do that is through having your stakeholders build a user story.

And again, this is something that Katie talks about all the time. A user story is a real simple three part Mad Lib as a your role, I need to a task. So that outcome, right, if you can have stakeholders or anyone fill out these mad libs, and you know, for every task that they have, it becomes abundantly clear what the purpose is, and how you’re going to measure the outcome.

As a content marketer, I need to write 1500 word blog posts in less than five hours each so that I’m not spending my entire day and part of my evenings writing content for my website. Right? Just with that statement, you go, Okay, well, our measure of success is going to be how much time it takes you to write a blog post.

There isn’t anything mentioned there about quality, but we probably might want to ask about that. Yeah, might want to say so that I write better ranking blog posts in less than five hours. it becomes clear what the task is, what the process is, right? It’s a writing process. We know who’s involved, which is you.

And suddenly, by writing out that Mad Lib, you understand how AI is going to play a role, right? Because AI is going to maybe do the first draft or the outline or help you brainstorm or maybe even do the second draft or do some developmental editing, whatever the task is within the writing process where you’re getting bogged down.

That’s where the machine’s going to step in and help you with that task. That’s how you, that’s how you achieve alignment. The very worst thing you can possibly do is start with AI, right? And say, let’s use AI for something like, let’s use a, let’s use a blender for something. Well, that’s a bit ambiguous, right?

If you’re making soup, sure. If you’re making steak, no. But if you don’t know what you’re making, it’s like, I’m just going to blend things. It’s not going to go well. You know, your goldfish might be very unhappy with you. So the. That’s the straightforward answer. Know what your purpose and performance is first, then figure out the other stuff and do the AI part last.

[00:41:56] Mike Allton: That makes a lot of sense. It was kind of how I approached podcasting. Those of you listening may or may not know. I now have six podcasts that I’m hosting and I was typically spending an hour or two preparing for each and every. Episode now with the help of AI bringing in research on the guests, it’s helping me craft questions, understand what they’re really going to want to talk about and what they’re excellent at.

This show is a perfect example. I’m using a custom GPT for every single show now to build out my pre show document, which I share with the guests. And it’s, you know, whittled that time down to a half an hour or less. It’s incredible how much time that saved, which I can now devote to finding more guests and all the other things that are important to my role.

But you talked about the cost and I want to touch on that because some of the people listening might be concerned about or hesitant about the cost and the complexity of integrating AI. What advice would you share with them?

[00:42:48] Christopher S. Penn: Again, it goes back to understanding the purpose and the performance, right?

What is four hours a week worth, right? What is your bill rate? If your bill rates 300 an hour, you can save four hours a week, that’s 1, 200 saved. And that’s that’s pretty good ROI. Even if your bill rates like 10 bucks an hour, right, there’s still 40 of time that you didn’t burn on a task that you had a machine do.

So that’s a big part is understanding what anytimes someone talks about ROI and cost, you have to take into account ROI is a financial formula, earn minus spent divided by spent. So you can either earn more or spend less. When it comes to AI, particularly the opportunities for the use of AI, it’s usually going to be on the spend less side, right?

So you can spend less time because time is time isn’t money, but time equates to money and people, people spend money to buy time. That’s what you’re essentially you’re doing with AI. So if you can use the tools to buy yourself more time by automating things or increase the quality of your work, then the cost almost is irrelevant, as long as there’s a clear financial impact on the earned side.

So for example, you’re talking about how you’d like to develop questions for guests, one of the things that the show hot ones does. Now they use humans do that I would not use humans to do this. But you can But they try to, they really strive to do deep cuts and ask questions that other people don’t ask.

One of the things that pretty much every show does is they kind of tend to ask the same general questions. One of the things that a clever podcaster with the power of AI could do is go to YouTube, grab every guest interview that you can find of that person, grab the captions files from all those videos, and then put that into a tool like Gemini and say, Oh, here’s the all the interviews that this person has done.

go through them all. What questions on this topic? I have this person not been asked. And you will come up with really good questions, insightful questions, challenging questions that will make the person go, Huh, no one’s ever asked me that before. Right. And that’s it’s a it’s an example of how do you use these tools to increase quality.

And so you now we’re on the other side of ROI. If the show is more compelling, the guest is like, that was a really good interview. I’ll spend some time maybe sharing that because I got to ask questions I’ve never been asked before. And for the listeners, it’ll be the Oh, it’s not the same old, you know, Oh, God, it’s Chris again, he’s gonna say the same eight things he said on the last 22 shows he’s been on.

Like this is now Oh, that’s different. That’s a question. No one’s ever heard before. That’s an answer. No one’s ever heard before. That’s pretty cool.

[00:45:38] Mike Allton: I love that. First of all, it’s a topic that Andy Crestodina, who was a previous guest on the show was talking about from a different perspective. He was talking about content gap analysis, and he made the point cleverly that as humans, it’s really hard for us to identify what’s missing.

Or what’s not their AI. That’s no problem whatsoever. And when it comes to podcasting specifically, I’ve done what you’ve suggested, right? I’ve done the research. I had Rory Sutherland on a show a few years ago and Rory’s amazing. His mind is, is intense. And. A wealth of information. We’ll put it that way.

And he’s been interviewed hundreds and hundreds of times. And so to find a question that he’s never been asked before would not only be interesting to your point to the audience, but again, to your point, would be incredibly powerful in building a relationship and rapport with Rory and I. Did that. And it took me days to sort out and listen and analyze all the different interviews he’s been on talking about agencies.

He works for Ogilvy who those of you maybe don’t know him. That was an intense project. It’s not something I could even remotely think about duplicating ever again. For most of the guests. When, as I said, I’m interviewing for six shows a week, but you’re right. What if I told you you could

[00:46:52] Christopher S. Penn: do that in 10 minutes or less?

[00:46:55] Mike Allton: Yeah, no, I’m serious. Like I

[00:46:57] Christopher S. Penn: told you, you can do that in 10 minutes or less, but absolutely do that. So let me show you, let’s do a quick walkthrough. Let’s go to, I’m going to share my screen here. Let’s go to YouTube. Let’s do Christopher Penn interview. Right. And we’re going to find people who are not the dead actor.

So we have a bunch of these things. So you would essentially go through and grab all these URLs. So let’s go through and I’m gonna just do a sample a few examples here. And I have a tool called OpenList, which just captures the URLs in your browser tab, right? That’s, that’s all that does. Very straightforward.

We take those URLs, we put them into good old fashioned text file. I’ll call this Chris interviews. Now, there is a free open source tool called YT DLP, YouTube Downloader. And what this tool does is if you’re familiar with the command line, it allows you to grab resources from YouTube. One of the things you can do is have it grab only the captions files.

So let’s put this in here. So it’s going to go through and grab every single one of these captions files and store them on the desktop there. These captions files are something that generative AI is perfectly capable of reading. So we’ve got a looks like we got eight or nine here. So let’s now take go into Google’s AI studio.

Let’s go into a new prompt here. we’re going to do some analysis of interviews. I’m going to provide you with VTT captions files. What do you know about reading VTT files. So we’re going to prime the model. I’m going to change the the pro model instead of the flash model because flash is dumb as bag of hammers.

I’m gonna turn off all the safeties because I like to live dangerously. And it’s going to read through the VTT files. Now the VTT files probably going to need to be converted into text. So let’s go ahead and do that. And we’re going to actually do we need to do that? I don’t know if we do. Let’s let’s give it a shot.

Let’s see what happens. We’re gonna go upload to drive. And let’s take one of the each of these little text files, drag them in.

It says text files we add from this prompt. Okay, it didn’t do that. So we’d have to add that in as change the format to text which got a few minutes here. I actually wrote a piece of code to do this because I got tired of doing it by hand sucks. And so there’s a captions.

All right, good. Now we’ve got text files. That was much simpler. We load those text files in

There we go.

It will do the analysis of them.

I would like to understand what questions Christopher Penn has not been asked in these interviews that are relevant to his field. of study, namely data science and AI. And so this is a toy example. You’d want to spend a little bit more time on these prompts. This is just a very simple example, but these are the, here’s what was asked and what did discuss here, some general topics that are at least relevant that were not covered in the Q and a things like he mentions RNNs, but does not delve into detail, et cetera.

So this is an example of, and we did this in what, seven minutes. Yeah, taking a bunch of YouTube videos, grabbing all the captions, putting them into Gemini, and the saying, here’s the questions you could ask as a podcaster that have not been asked. And you could take a little bit more time to find more interviews, maybe find some podcasting news, get them transcribed and so on and so forth.

This same technique for marketers, by the way, is one of the things that you should be doing with your marketing. Because guess what? How many, how many really good sales podcasts do you know of Mike?

[00:51:18] Mike Allton: Sales podcasts. That’s not my area. So maybe a couple,

[00:51:22] Christopher S. Penn: right. Well, you know, they’re out there. There’s like a couple of dozen, like really good sales podcasts, folks like Marcus Sheridan, Matt Hines, et cetera.

Go grab the captions from their sales podcasts, have it build a comprehensive guide to how Marcus Sheridan thinks about sales. Now you’ve got a, a blob of knowledge says this is how Marcus does sales. then take that and say, here’s our sales playbook as an organization, vet it against Marcus’s methodologies.

Tell me how we could improve our sales. If you’re a marketer, go to Andy Crestodina’s channel on YouTube, where every interview Andy Crestodina has done, grab all his stuff, digest it down, or go to his blog, which is amazing. Grab it, digest down, say, here’s Andy’s tools, you know, ideas for how to event a landing page.

for conversion rate optimization. Here’s the HTML for our landing page. Tell me how to improve it using Andy’s ideas. Every single piece of content that you have ever come across as a marketer, an ebook, a white paper, a guide, a podcast interview is fodder. for an AI tool to create rubrics and scoring mechanisms and analysis of your stuff.

So as a marketer, you can sit there and say, well, I’ve got this email campaign, who do I know that does amazing email marketing, grab all their stuff from pub in public, right? Don’t, you know, use stuff that’s publicly out there. They’ve they’ve put on the internet for you to consume and say build me a outline a summary of the way they think.

And now vet my content against that. So you can take all the stuff that you know, you should be doing that, you know, you should be doing more with CRO, you know, you should be doing more with email marketing, you know, you should be doing more with SEO, but you don’t have time, take all this knowledge that you have captured for the last 510 1520 years of experts and have AI pretend to be that person and apply it against your stuff, right?

That is the smartest thing that you can do with the remember, whoever has the most best data wins. If you take the time to accumulate all this data that you’ve have been with all these conferences, you’ve gone to buy all these podcast interviews you’ve done and turn it into that AI can effectively leverage, then you can use the benefits of that.

My friend Tamsin Webster wrote an amazing book called find the red thread. She’s done a gazillion and a half speeches. I’ve taken the public speeches, not her book, because that’s not given away for free. But the speeches she’s done on YouTube that she gives away for free. I just did that down said, is still the essence of the red thread framework.

It does the AI does that says, Oh, now I’ve got my virtual Tamsin. Every time I have want to have legs say here’s my content, virtual Tamsin, how would you apply the Red Thread framework to this and make it better? And it says, here’s the things that you’re missing. Less nicer than Tamsin does. Tamsin is a wonderful human being.

Like, oh, I forgot this and this and this and this. And so I have a virtual Tamsin anytime I want to vet my content. So the people that you look up to that you admire, In marketing, in sales, in whatever discipline use the publicly given away stuff, and you have that expert on tap whenever you need them.

[00:54:43] Mike Allton: Wow. What a powerful way to finish Chris. That was incredibly insightful folks. I think you’re going to want to go back and listen to that segment several times and really just ingest this tactic and this strategy that Chris just shared. That was fantastic. Chris, for folks who do want to learn more, who want your help and want to bring you into their organization, where can they go to find you and learn more about you?

The easiest

[00:55:06] Christopher S. Penn: place to find me is a trust insights. ai. Or if you want the personal version, go to Christopher s pen. com. But trust insights. ai is where you can get our professional help called out operators standing by.

[00:55:22] Mike Allton: Fantastic. That’s all we’ve got for today, friends. Thank you, Chris, for being with us.

Thank you all of you for listening. Please don’t forget to find us on Apple, the AI in Marketing: Unpacked podcast and leave us a review. I’d love to know what you think until next time. Thanks for joining us on AI in Marketing: Unpacked. I hope today’s episode has inspired you and given you actionable insights to integrate AI into your marketing strategies.

If you enjoyed the show, please subscribe on your favorite podcast platform and consider leaving a review. We’d love to hear your thoughts and answer any questions you might have. Don’t forget to join us next time as we continue to simplify AI and help you make a real impact in your marketing efforts until then keep innovating and see just how far AI can take your marketing.

Thank you for listening and have a fantastic day.

Related