Andy Crestodina, Co-Founder & CMO of Orbit Media, joins our host, Camela Thompson, Go-To-Market Thought Leader and B2B Insights Expert, in this episode of the Revenue Marketing Report. Andy shares how to use AI for content marketing in B2B.
It’s so funny because if you go outside the marketing world and start looking for articles on AI, it is pretty doom and gloom. So I am happy to be talking about it in a positive light. It’s going to be a differentiator for those of us who embrace it and learn how to be more productive with it and those of us who run away from it. It’s great!
“Yeah. I am happy to have this conversation about bias and societal impact, and the labor market. I’m trying to get value out of the tools I use and I’ve just a purely practical perspective when it comes to AI while I am working.”
I’ve cheated and heard some of your pro tips, but would you please share with our audience what you found working with AI in your content strategy?
“I am going to start with two things. One is our goal for using AI is for greater productivity and efficiency. I’m going to set a benchmark. I want to be 20% more efficient. That is it. I am not asking for anything else and I would be amazed with that outcome. That would be a big boost.
“Secondly, my audience, my goals and my voice are different from all other brands. Your audience is different than every other brand. We are all different. Therefore, with that in mind, I highly recommend that we start by telling AI a lot about our audience. Every strategist in the world would never just begin by doing execution. You have got to stop and think about it for a minute; who are we talking to? What are their information needs?
“I like to start with a persona prompt. If you’ve clearly documented ideal client profiles or battle-tested personas, you can simply upload them. That is the ideal. If you don’t have those, you can begin with a prompt like this:build me a persona of a job title at a company of size X in industry Y in geography Z with these sorts of business challenges and goals. At the bottom of that prompt it starts to sound like: tell me their hopes and dreams, their fears and concerns, their emotional triggers and decision-making criteria for selecting a company in my industry. What you eventually get will probably be named Alex or Taylor. It will make a persona for you.
“That persona will be incorrect and that’s fine. Improve it, fix it, and edit it. Ask the AI to remove something or add something. You don’t have to trust it. We are not supposed to trust it. You are supposed to take what it generates and improve upon it. Then we have a persona that we believe in. This feels like we know our audience and now the AI knows our audience too.
“The next step happens in that same thread. You’re going to keep talking to the AI about what the topics or concerns are. What do they care for? What would they find unexpected? What are they hoping for? What is the most useful thing I could publish? What are the tips that could be the most valuable for them? What research might they click on? What key phrases do they search for? What social posts might they click on? And it’s all the way more interesting. You can run your sales script against it. You can upload your PowerPoints and ask for feedback. You can hit it with your sales day or night. It’s very, very valuable to have synthetic versions of your audience and you can obtain insights into how much money you need to spend on this. What is your decision-making process? Do you have six buyers or one or who do you trust? Who are you listening to? What podcasts do you subscribe to? Literally, anything you need to know. Then, you can work on it and begin to get insights on that.
“The specific content strategy prompts come next, but they’re essentially right along the lines of what you would expect; frequency, formats, topics, promotion channels, and other things that aren’t specific to the audience. However, you could also put all that together into your plan for your content program.”
When I hear people say I have tried AI to write copy for X and it was terrible, the first thing that comes to mind is, did you know that you can have feedback cycles and keep iterating? So what you described was 20 steps better because you started with the persona and what they cared about. But you also mentioned it is going to be wrong and that’s fine. Of course, tell it what is wrong and fix it. Sounds simple.
“If you hired a person and then gave them one sentence of instructions and they didn’t automatically get an A plus on the job, would you blame that person? It is not their fault. So it is way more iterative, but it takes way more work. It’s harder like anything else. It’s strange that people type one or two lazy prompts and then judge the entire technology as being incompetent. In the end, these are things you are building up. You’ve to train it. It’s very tedious and time-intensive.
“There are two types of training that I recommend. One is to train it on the audience. The other is to train in it on your brand voice by giving it 55 of your articles, tons of your headlines, and lots of your social posts. Export things and give them to it. It will start writing in your voice and if you train it in detail, it takes like a hundred X more training materials than people think. But in the end that is where you’re going to get value out of it. Then you’re well positioned to get that 20% productivity boost. I’ll recap. You’ve got a persona. Then, ask the persona for top concerns. Turn those concerns into article ideas. Turn those ideas into creative briefs, drafts or outlines and then optionally, that is when you’re ready to ask it to write an article. I am already eight prompts deep. It might have only taken me ten minutes to get there. However, I don’t think that that long shot prompt requesting it to write an article about X is just never going to work.”
I have learned the hard way to stay in one conversation lane, but you mentioned a great tip about asking it to write a style guide. Have you done similar things with personas where you ask to summarize that so you can just feed it back to it and get something concise?
“Yeah. if you have it write something for you and it wasn’t good and you took it out of the AI and you fixed it, the magical final all-important step is to feed the final draft back to the AI so that it can learn what you just did. You’ve got a better chance of it giving you a better draft next time. Overall, we are going to be building this up and if you want a style guide, that’s very tedious. I have done it. It took me an entire Sunday morning. I copied and pasted 60 articles into a single Word doc. Now I have a massive Word doc with two years’ worth of content. If I upload that to the AI, I need to give it a prompt to just acknowledge that it has got it. So it sounds like this:
“I’m giving you two years of Andy’s articles. Summarize his style or write a style guide based on that knowledge base so it will give you back something. If you haven’t created a style guide, asking AI to help write prompts is another magical trick. That style guide is something you can share with your team. You can give it to other people. You can use that. You can keep on improving that. Keep it separate. If AI forgets because of the size of the document you uploaded, you still have your style guide that you can load in. And then, there are people who are turned off by AI since it utilizes a lot of AI-typical words. It always wants me to write headlines that run: unlock secrets. I would never write unlock the secrets, that just sounds like spamming. So weird. I also don’t write unleash. I would never write that. Those words are typical AI. So make a list of words to avoid and upload it. If it keeps using a word you don’t like, just add it to the list. It will stop using it forever.”
The feedback loop is important and like your point that if something isn’t working, tell it and it will adjust. I do want to talk a little bit about a bias that might not be the worst thing and that’s recency bias for companies who are just figuring out their product-market fit, and their personas. How often do you think people should reassess their content strategy or what are the signs they need to look for more recent customer information and adapting?
“That data is almost available in real-time in your analytics. Social media is a data-rich environment. It’s hard to export social media data and feed it to the AI. It tends not to be very structured. I struggle with that on LinkedIn. You may not be thinking of it this way, but everything you post is in fact a bit of an experiment with feedback and you can start to see these things get great traction, those that got bad traction. The sales conversation stuff is a tough one.
“I don’t have a great answer. Your sales closing rates are down or perhaps you don’t have good answers to people’s questions. You are finding a common objection or a new objection that you weren’t expecting or haven’t supported with any articles yet and depending on how fast your industry moves, that could be every couple of months, I think that industries have speeds and some move very fast like design and technology industries. For other industries, the questions have been the same for twenty years and you’ve got to nail them, but they don’t change that often.”
And sometimes, you may be writing things that are appealing to a certain audience. So your page visits and your engagement rates look good, but not your conversion rates to actual demos and other bottom-funnel things. I think if there’s a mismatch there, then it is probably a sign that you need to reevaluate something about either the topics you’re focusing on or the persona might be a little bit off.
“Yes, it happens, especially right now. The pace of change is different from what it once was. Things are happening much more quickly. There’s a big disruption. From GA-4 to AI and now pickleball So, it’s a good time to reassess your personas.”
Oh my gosh! That’s so funny. I remember going to a park recently and I wondered what are they doing? That felt recent, but it was probably a year ago.
“Yeah. when you see people at the park doing something and you don’t know what it is, get out your personas. It is time to take and look and revisit those.”
Or if people start calling the market something like SaaSacre or the diSaaSter, it’s probably time to reassess.
“Do you remember when the word cloud replaced the word internet? It’s like where did that come from? It was just one day and everybody decided that, like SaaS and freemium. So AI isn’t recent enough. You can only try. You have to. What is worthwhile sometimes is to scan the titles of sessions at major conferences. Those people spend a lot of time trying to catch lightning in a bottle and those may give you clues on what’s trending or about to trend.”
But what’s funny is they are usually doing calls for speakers a year or six months in advance. So you’ve probably missed the boat anyway.
“I am not a fast adapter, people think I’m not that early on anything. But when I see things come and I skip huge trends. Clubhouses, what is that? Blockchain, web 3.0, NFT, I’m paying no attention at all. However, AI was one where I saw it’s worth blocking time on your calendar for an hour a week to just run experiments.”
I think that’s fantastic advice, very smart and I hope people get a lot of value out of this and start using AI to become more productive.
For more content on B2B marketing trends, listen to the full Revenue Marketing Report episode at the top of the article or anywhere you podcast.