Can AI content break through?
Tom Swanson
Summary
The keys to effective use of AI in content development are: spend the time to develop your positioning, feed your docs and data to the machine, work on your brand voice, and give it a recursive learning loop so it can learn from each post.
By Tom Swanson, Senior Engagement Manager at Heinz Marketing
Recently I was talking to a past client about how he is using AI to generate content and the big question I had was: does it work? Does it actually break through or does it just wind up sounding like everything else?
I was surprised to hear an emphatic yes.
So I asked him more about how he has made that work for him, and here is what I heard and what I think B2B marketers need to take away.
As a quick aside, it seemed ridiculous to write this post with AI, so this one is all me.
AI content isn’t an inherently bad thing. We have a few posts about it:
- Does your AI Output Feel Generic? – My post is a follow-up to Lisa’s, to share a story about how this has worked for a client
- Agentic AI and Content & Messaging – This is a guide for revenue leaders
- Agentic AI and B2B Metrics – How to measure these programs
This has taken me a lot of time to get over. I won’t spend too much time getting into it here, but I have come to the conclusion that there is good and bad AI content just like there is good and bad human content.
If you are against AI content no matter what, then this probably isn’t going to change your mind, and frankly, more power to you. What remains, though, is that the same things that make for great human content also make for great AI content, so keep reading if that is of interest to you.
Marketing 101
AI content that performs is built on a solid set of marketing foundations. It is as simple is that.
The most important thing in marketing today is: positioning. Same as it ever was. Do you have a position that you can stand upon confidently? Do your “uniques” simply erode away when your competitors show up to the party?
You need to go take a look at your core foundations and re-evaluate their uniqueness against the rest of the market. There may be enough space online for everyone, but there isn’t enough space in the mind of the buyer for every vendor.
Furthermore, LLMs favor positioning that has something unique to say and an interesting way to say it.
This was the single highest-value thing that this former client cited for their AI-content program. They put a major focus on their positioning up-front (it was a big part of our work together, shameless plug) . That work paid off as they are now able to reliably produce brand-aligned, valuable content that utilizes their highly specific and well-developed market position.
So dust off that old marketing textbook that talked about such dated concepts as “knowing your audience”, “having clear value propositions”, and “making good offers”. That stuff is the key to success in modern marketing. Having some real déjà vu.
Feed the Machine
Arguably the most important thing you can do is to feed this machine from your sales data. There are all kinds of sales tools that gather and aggregate data, and most of them are just one API-call away from your content engine. This is likely the highest value thing you can do for your content.
Guidelines and rules are crucial here to avoid any sensitive data getting out. It would follow, then, that you should have a separate layer in any agentic content system that gathers data, analyzes it, and pumps out trends and useful insights without including anything sensitive.
Of course, this should be coupled with human review for safety.
Information inflation
Full disclosure: I love this concept. Information Inflation is something I have been thinking about for years. I will spare you my longform musings.
Information is so easily accessible and, equally as important, so easily produced. Now with LLMs, it is a safe bet that any point in time, everyone knows everything.
Every single piece of content simply has lower marginal value than the one before it. No matter how brilliant your content folks are, it’s tougher than ever to stand out.
From what we discussed, there are two ways we considered to handle this.
The first is to have an interesting voice. This goes back to the positioning discussion of making sure you are set up to give your AI tool what it needs to make content that works. If you want to be snarky, then teach it to be snarky (careful with this one). If you want to present things pragmatically, give it lots of data to use. If you want it to be realist, then be honest about your market field.
The alternative is consistent volume. Just like the fed prints money, you can print content faster and cheaper than ever. I don’t care for this one, but it is effective. Fresh content is better than stale (especially now), and a robust content body means more data points to learn from.
Recursive Content Generation
A bit of a misnomer, but it’s a catchy section name.
If you want to get agentic with it, and you should, you need a loop-closer in your content system. Essentially, this is an agent that pulls content performance data (whatever the format) and feeds it back in to the writer.
Your content engine sees what works and what doesn’t and then can update its operating system to output better content for the audience. This is the last piece.
It is shockingly easy to build these sorts of agents. If you want a quick start guide, here it is:
- Get Claude Code
- Make a folder on your computer
- Initiate Claude into that folder via CLI (I like doing it this way, makes me feel like a hacker)
- Ask it to plan out an agentic content engine
- Work with it on development and test
This is how I have been building out agents, and its been pretty effective. While I wouldn’t recommend going full bore into producing apps, using these tools to assist with data collection and analysis, then feeding that back into your content generation is a solid move.
Conclusion
There are so many ways to make content these days. I would recommend exploring as many as you can. Video is a particularly interesting one that I have little experience with but a big interest in. Video content generally outperforms most other formats, so I would be curious to hear from anyone who has had success automating this particular format.
Anyway, I wanted to share more about this conversation, as the use of AI in content is something of a bogeyman amongst marketers. I don’t think it is all bad.
I am not without concerns. I do worry that content becoming increasingly commoditized is a net negative, and that it will result in poor overall quality, but this is hardly a new concern. It just is accelerated from AI.
Of course I worry about removing humans from the process as well for many reasons.
Ultimately, I can’t say where the content world goes next, but it is clear that this a powerful tool when used right. It isn’t something to be shunned, even though it is easy and trendy to do so.
If you want to talk more about this, or better yet, how you can build up your fundamentals, we are just an email away at acceleration@heinzmarketing.com.
