Generative AI Doesn’t Make Content Marketing Easier, Just Hard In a Different Way

CEO Corner

It’s clear that marketing will go all in on generative AI. The opportunity and potential have been well documented. In a recent meeting with my AI team at Aprimo, I asked, “We understand what will get easier. The question is, will anything get harder?” I was amazed that in a few short minutes, I had documented over a dozen ways content marketing will be more challenging in a generative AI world. I’ve highlighted a subset of those to get people thinking about how their marketing organizations will have to evolve to truly take advantage of a generational technology shift.


Working for the Digital Asset Management leader, I see large enterprises and digital-first/digital-heavy organizations struggle to organize and leverage the vast amounts of content that they create today. I wonder what those challenges will look like with one or two orders of magnitude more content.


There is now a ton of prompt management work to be done to articulate market segment definitions, brand guidelines, and writing style/brand voice. While I don’t think “prompt engineer” is a long-term role, getting the team up the learning curve and managing the brittle phrases that produce the best results is a hassle and doesn’t scale.

There are also a lot of new training data challenges for models. Getting fine-tuning or embeddings to encapsulate your product knowledge, messaging framework, and other data sources is hard.

Budgeting Complexity

Optimizing storage costs and associated processing costs of content is a reoccurring theme. I wonder how many budgets for 2024 are contemplating that storage number increasing by a factor of twenty…or two hundred. 

I also wonder how large-scale utility token pricing/consumption is going to be managed. When cloud first started to go mainstream, there was so much budgetary carnage from unexpected utilization overruns. Some poor developer’s unit tests ran wild over a vacation, and the enterprise had a huge budget overrun. Now, this is a core function within DevOps. I wonder if instead of a long-term title of “prompt engineer,” we will have a marketing operations title of “token overlord”?

Brand Safety and Compliance

People are terrified of AI turning into a monster, so brand safety analytics becomes more important and must be moved into the review process. While that may be an overreaction, getting generative AI on brand for text content is pretty hard and virtually impossible in scalable/practical terms right now for image and video. Lots of vendors are stepping into the void here, but that, in and of itself, tells you there are hard problems to solve.

There are also basic quality issues (which can be overcome with tons of prompt and training work). A lot has been written on the quality of output and the work that has to go into improving it. When you hire a copywriter or other creative, they get better over time as they ambiently absorb information about the brand. With generative AI, this process has to be explicit. Tracking content performance, feeding higher performing content back into training data, tightening metadata, experimenting with prompts…in the new world, what happened naturally now must be an explicit continuous improvement process.

Democratized and decentralized content production removes the bottlenecks that acted as effective safeguards for organizations. When content was created by a relatively small number of highly skilled and largely centralized resources, it was easy to inject governance processes into the content creation process. In a generative AI future, those safeguards will have to be subsumed into a new or revamped content operations infrastructure and pushed closer to the edge, just prior to channel applications like CMS, social media management, and marketing automation.

There are also the much talked about and more ethically murky issues like training data copyright and bias concerns. I’m not sure if that is any different than the last thirty years of the digital era, but because it is so much more explicit (and perhaps traceable), it creates more liability and process hoops for organizations to consider. I’ve had hundreds of conversations on generative AI and marketing. They absolutely inform my opinions. When someone has a lot of experience, they are called an expert. If it is an algorithm creating the derivative IP, it is viewed a lot differently.

What’s Next?

I think we all know how exciting generative AI is. It is the biggest innovation in content since the internet, radio/television, or the printing press. But, I think we have a lot of things to overcome. I would love to hear what you are all doing to prepare and lay fundamental processes, strategies, and process infrastructure to take advantage of the generative AI promise.

Have you heard about our new ChatGPT integration? Check it out here.

ChatGPT Integration Demo CTA 1

Don’t miss a beat!

Sign up to receive our latest content on best practices, trends, tips, and more to elevate your content operations.

Don’t miss a beat!

Sign up to receive our latest content on best practices, trends, tips, and more to elevate your content operations.