Creating an AI Automated Marketing Department: How AI Agents, Skills and MCP Are Transforming Marketing

0:00Welcome back to the Digital Marketing Podcast brought to you by TargetInternet.com. And in this episode, we're talking about creating an AI automated marketing department.

0:17Okay, before I get going, I should start by saying that I'm not saying that all marketing should be automated. Don't need marketers or the AI should be allowed to create content and publish it without a human in the loop. And in fact, if you've been following me on LinkedIn, I got myself into a lot of trouble recently. When not so long ago, I said you can replace a marketing exec with three prompts. And here are the three prompts. And LinkedIn promptly lost its mind. Now, my point was rather that we need to start reimagining these roles. Because actually, if the job is to do some research, create some content, post it online, create some accompanying social media posts in various different image sizes, post those online, review the analytics data and iterate.

1:01That's all stuff that AI is really, really good at. So we need to start looking at what we all do every day in our jobs and start to think about how we can take away the lower value, the more administrative stuff and focus on where humans really add value, which is creative, strategy and having taste and empathy based tasks. things like understanding our audience and building campaigns that resonate now when i look at which roles are being most impacted by ai two that always come out are developers and marketers and whenever i show that to my marketing students they look at it with horror but what i actually say is that there's a huge opportunity here because if you're really good at marketing understanding your target audience working out what resonates with them what they value, what's important to them.

1:51But also you're really good at AI. There's a huge opportunity for combining those two things really effectively. And marketing is a very human thing. It's about understanding what motivates people. And it's about really trying to get underneath the skin of why we like something, what we like, and trying to resonate with real human beings. Now, as you might know, target internet members, get these half-day masterclasses every month. and this episode started based on the most popular one we've ever done. So we did a half day on creating an AI automated marketing department and went through all this stuff step by step.

2:27But what came out when we were building that was some really interesting key techniques and frameworks that I think are absolutely game-changing for marketing. So I wanted to share them here. Now, by the way, Target Internet Membership is free for a month still. So if you want to get in and try out the masterclasses, go over to targetinternet.com. You can sign up, you get those masterclasses, those one-hour update sessions every month, as well as that big library of other stuff. Anyway, to give this whole piece some context, let's just start by reading through a prompt that I used recently that first seems pretty simple, but actually when you untangle it, it shows something fairly profound.

3:07So this is the prompt. look at my web analytics using the Windsor MCP and consider my current content using the target internet MCP. Research and write an article on an analytics related subject that's in demand, use the target internet brand skill for tone and diagram styling, make the article live, then create a carousel post for LinkedIn that summarizes the article and links through to it and schedule it in Metrical for this coming Monday using my connector. Okay, sounds pretty simple, but let's really go through it and untangle a couple of key things in here that are really game-changing. So the first line, look at my web analytics using the Windsor MCP.

3:49MCP, probably one of the most important things we'll talk about today. Model, context, protocol. Think of it as the kind of USB-C for software, so you can connect it up to AI. So if your particular piece of software, your content management system, your customer relationship management system, your analytics or anything else has MCP, you can connect it up to your AI of choice. And then essentially you can retrieve information, but you can also write information potentially. So in this case, I've used something that allows me to connect to Google Analytics and Google Search Console, which is a brilliant tool called Windsor.

4:30Now there's two types of MCP servers. One is an online one. So for example, in the moment, I reference Metricall. Metricall is one of my favorite social media management platforms. It's really low cost. It allows you to build your social media posts, schedule them, get analytics, all that kind of good stuff. But it's got MCP and it's an online server. So you just go to your AI of choice and you can connect directly up to it. The other type of MCP server is what Google Analytics has got. and this is when you need to set the server up yourself and host it on your machine and it's actually very technical and it's not something suited to most people.

5:10So therefore in this case we're going to use another tool to allow us to connect to that easily and that's something called Windsor. So Windsor.ai. The free version allows you to connect one thing permanently so you could connect up your Google Analytics. I use the Next Account Up which is a bit less than$20 a month and that basically allows me to connect up a number of tools. So I'm connecting Google Search Console, Google Analytics in this case. Then I go on in my prompt to say and consider my current content using the target internet MCP. Well, that's just for our content management system.

5:47So most of the big content management systems like WordPress now have MCP so you can talk to it via your AI. So I've asked it to basically look in my analytics and look at what current content we've got and work out based on some research what might be in demand in terms of an article about analytics. I've then gone on to say use the target internet brand skill. Well we covered skills in the last episode but essentially a repeatable way of doing something. So we've got a skill so it knows our tone of voice, our colors, our layout, all those kind of things. So it can write in our tone and it can create diagrams in our style.

6:25Then I've done something you should never do and I said make the article live. So I've got it to write something and to post it live online without even looking at it. That's not a good idea, but it was for the sake of a demo. Then create a carousel post for LinkedIn that summarizes the article and links through to it and schedule it in Metrical. So again, I said, Metrical has MCP. So I can, I read information, but I can write information. So what did it do when it went off and it wrote a brilliant article? And that article had lots of data in it that was unique to us because it had looked at our analytics and used that as an example.

6:59So if I think about search engine optimization and I think about Google's EAT, which is their experience, expertise, authority, and trust, well, that's demonstrating expertise because it's an in-depth article with lots of good referencing. But it's also demonstrating experience because it's got unique data in there. It's not commodity content, which is something that Google are talking about a lot at the moment. So it's created a good article. It's got diagrams, original research content, bibliography, all that kind of good stuff in there as well. And it's been able to post that online into our content management system.

7:35But then it's been able to create a carousel post. Now it created a carousel post that completely followed our brand guidelines, but wasn't actually how we normally create carousels. So I told it that. I said, actually, here's an example of a carousel we normally create, to which it said, no problem at all, I'll update the brand skill so when in future you ask me to go through and create a carousel, I know the way in which to do it. So if you think about this, it's looked at analytics data, it's looked at search console data, it's looked into my content management system, it's done some research, it's created something in my tone of voice, in my styling, it's been able to upload that to my content management system, it's been able to create a social media post in the right format and it's been able to go and post that online.

8:21So that single paragraph prompt is really going through and doing probably 80 % of what a marketing exec does every day. So we really need to start looking at all the things that we need to do. But really, the combination of skills, repeatable ways of doing something like our brand and MCP that allows us to connect to external tools, we can really start thinking about automating or semi-automating a whole range of tools. If you add some scheduling to that, so it could do it regularly, I could take away a lot of the work that I'm doing every week. Now actually, in fact, tools like Copilot Studio actually allow you to trigger agents and prompts when something happens, like an email arriving from a particular email address or a task be added to your planning tool.

9:10meaning we could create these automations that happen autonomously in the background without us even being involved in many cases so given this context of skills and mcp let's think about the maturity of the organizations we work in and their use of ai as well so if we look at a maturity model and you can find this maturity model in the show notes so targetinternet.com forward slash podcast. We start ad hoc where individually we're using different chat and kind of talking to AI and doing things, maybe drafting some copy and coming up with ideas. Then we hopefully move towards standardized, which is we're sharing prompts.

9:52Maybe we've created some brand skills, maybe there's certain tools that are approved that we can connect to. Where we're suddenly getting to now because of MCP is connected. And that's where we can read and write from our content management system, our customer relationship management, our ad platforms, our analytics, and so on as well. That allows us to move to stage four, which is automated, scheduled reports and workflows that are actually running without us prompting. And what we eventually get to is AI-assisted teams, where we start to redesign our roles around the different automations and agents that we've created with human sign-off in the loop where it's appropriate.

10:35So if you think about it like that, the next stage of this is for us to start to reimagine what we do every day. And therefore, we need to start looking at what we do do every day. So to do this, we need a kind of structured approach. We can't just automate everything all at once. It's very likely to be an iterative approach. So I wanted to talk you through a framework that will help you do just that. And again, you can download this in the show notes at targetinternet.com forward slash podcast. But what we've really got here is a quadrant. If you imagine along the bottom, we've got the amount of judgment and risk involved from low through to high.

11:13And on the side, we've got how often the task happens going from low through to high again. So let's work our way around the quadrant. So top left, which is high frequency, low judgment. So maybe running your weekly reports, tagging up lots of images, resizing, doing the first drafts of social posts or pulling data in. Those are all things that we could very much fully automate. Then top right, which is high frequency, high judgment, writing campaign copy, emails, budget changes, replies to customers. Those are the kind of things where we'll get AI involved, but we're going to very much sign those off.

11:53Then bottom left, which is low frequency, low judgment, well that's ad hoc. One-off landing pages, brief meeting summaries, that's stuff that we can do using the tools that we've got around us all the time already. And then bottom right, lower frequency, higher judgment and higher risk, creating strategy, positioning, crisis management. That's the stuff where we're going to lead human first. We might use AI for research. So what I would suggest that you do, you look at this quadrant probably a good place to start is top left and top right so high frequency low judgment and high frequency slightly higher judgment if we start with the top left high frequency low judgment we can start to think about what we can automate what can I take out of my weekly work and actually by combining MCP skills I can start to test this out and I can work out what's working what's not, can start to free up time.

12:48Some of my advice is sit there and write down with your team what you do all day every day and then start to look at this and work out how you could build out those automations. And actually if you start with the high frequency low judgment stuff, you can run it and you can see is what you're getting back accurate. Now with things like weekly reports, what you're doing is making sure the data it's pulling continues to be accurate. And if you're getting it to do any analysis of that data, ask it to state its confidence. Ask it to state where it got that data or that insight from. But start like that and then build up gradually and you'll find that you're actually freeing up time.

13:28And then we can start to lean into, well, actually, where do we add more value? And I always come back to, are we just trying to do stuff faster? Is it about just doing more stuff? I'd hope that's not the case. Is it about doing better? Can we do deeper research, higher quality content? Or actually, could it about be doing something completely new? Could we start to reimagine our strategy, create interactive content, start to do things that we haven't done before? Because that's what freeing up this time really looks like. So if I go back to where I started, which is actually about how do we reimagine our roles?

14:03How do we look at what a marketing exec now does? We're still going to need the core skills. Because if we're getting AI to do something for us, we're going to need to check the quality of that. We're going to need to check the accuracy of it. And if we don't understand that in the first place, then we're not going to be able to double check the work it creates. But actually, once we get past that stage, we're going to be able to work on the more interesting stuff. But if we think about our jobs, who's in charge of creating the skills? Who's in charge of creating the automations? Who are the people that are signing things off as we go along.

14:36And actually, who's coming up the strategy in the first place? So we start to move from very channel-based roles. I'm in charge of SEO. I look after social media. So actually thinking about value and judgment-based roles more as well. And that means we can gradually start to reimagine our marketing departments and can build as much automation as we can. But we need to decide where we are using AI and where we're not and where we really don't want to. I think that's really important. Now, what's quite interesting about all of this is it's not because the AI models have suddenly become so advanced that we can suddenly do this stuff.

15:12That's part of it. That is enabling it. But it's really four things. It's one, the agentic nature of AI now, the fact it goes in and solves problems autonomously. But secondly, it's about skills. We have this standard that works across pretty much all AI platforms where we can create things in a repeatable way like our brands. But then we've got MCP, Model Context Protocol, which means we can connect up to all those different external tools. And then finally, we've got these co-work or work models. So if you think about co-pilot co-work, clawed co-work, chat CPT, it's just called work. but essentially the fact that it can manipulate files on your machine which means you can do things at scale so instead of it all being in the context window in this memory it can actually say well I've got a thousand things to do so I'm going to create a text file markdown file that lists those thousand things and then work through it which means that you can have tasks running in the background you can work on much bigger scale projects those four separate things have come together at pretty much the same time and a lot of this wouldn't have been possible six months ago But now it means where the competitive advantage lies is if you can embrace using those different things, you can look at that quadrant and say, what do we do every day?

16:34We can start to take the lower value tasks and automate those. We can free up time and we can free up the time to either work less. That would be lovely. but maybe to do more creative things and to do more original strategies and really to work on the stuff that's always been of the highest value but quite often gets pushed back. So it's a really interesting time to look at this and I think it really is an inflection point with AI where we suddenly start to do really practical work and we start to look at actually how does work get done and start to reimagine that. So get over to targetinternet.com forward slash podcast for the maturity models and that quadrant that allows you to reimagine things.

17:15And we'll have more episodes coming up soon covering all of these topics. So thank you so much for listening to the Digital Marketing Podcast. For more episodes, resources, to leave a review or to get in contact, go to targetinternet.com forward slash podcast.