Generative AI is in the midst of having a major moment in the content management space. One of the long-standing hurdles for any organization is creating large amounts of high-quality content – particularly as the number of audiences and channels continues to grow. Generative AI has the potential to dramatically reshape how marketing organizations create and deliver campaigns. As such, it is now a major evaluation factor in platform selection – being top of the list of criteria from the latest Forrester CMS wave and vendors are starting to use it liberally in nearly every external communication.
(I will note that most pieces about AI from vendors are vague and heavy on hype – but Michael Andrews of Kontent.ai does a very good job of going a level deeper in most of his writing.)
On the flip side, Generative AI is literally at the very top of the Gartner Hype Cycle for Artificial Intelligence, 2023 – so what does all this mean?
we overestimate the impact of technology in the short-term and underestimate the effect in the long run.
Amara’s Law – Roy Amara
Amara’s Law is useful in helping us understand the timeframes for how generative AI will change how organizations work with content – and what to expect.
Short-term
Generative AI tools have already been incorporated into existing content management systems. Most vendors have been quick to develop plugins to assist with common tasks – such as generating titles, translating content or creating content models.
In other words the short-term effect is to enable the current tasks of content creators, but do them faster (and ideally better).
When thinking about the content that most organizations need to create, Jeff Eaton of Autogram has a great Mastodon thread about generative AI, “Content as Communication” and “Content as bait” – and to some extent most marketing organizations do need a balance of both. I would consider myself a decent writer and communicator with a specific voice, and as such would often be completely disappointed by most of the AI attempts to out-write me on any content that needs to be fairly novel or clever. It’s often bland and full of hallucinations, meaning that it’s best used alongside a human who can evaluate the output carefully.
That said, we regularly use tools such as ChatGPT to accelerate “content as bait” – because no matter how much you consider (and respect) your core audience of readers, the sad fact is that before you consider any of that, you do need to get by that initial gatekeeper – Google – and having the “right” content with the “right” SEO keywords is key to capturing that initial interest.
It’s this latter case that AI does well at, but it’s often being pitched in cases where a high degree of accuracy and, dare I say, actually compelling writing is needed. Certainly iterations of prompt engineering to properly convey all the information (brand tone, goals for the piece, underlying references, etc.) to generate high quality output can be done – but more often than not, a content creator will simply do the majority of the writing and simply use AI to smooth out the edges. In other words, the majority of what we are seeing today is an evolutionary, not revolutionary change to how we create content. And, as per the Hype Cycle, the gap between the hype and how organizations are actually using generative AI will lead to some deflated expectations.
Long run
That said, in the long run, I see the potential of AI to completely change how we think about content management. We have seen some uses of the term “content as data” – but currently content itself requires too much human intervention around content placement and contextual awareness to truly make that happen.
In order for this revolution to take place – the content creation tooling itself will have to massively evolve. To enable the type of mass operational scale we are talking about, instead of working with individual items of content, we’ll need new paradigms for managing the “sets” of content – which may include managing hundreds of variations along different axes; audience, language, region, channel.
An example that comes to mind that is not possible to do at scale with the existing paradigms is to imagine personalizing a since piece of content with 10 audiences, 15 languages and sub-locales (Québécois French is not the same as Parisian French, par exemple) and 4 channels (web, mobile, text, email) generates 600 different (but related) distinct variations. Now add the governance to ensure accuracy and brand voice – not to mention the ability to apply A/B testing to immediately retire and re-try newer and more successful iterations.
In other words, in the same way that the industry has moved beyond managing the page layout and copy towards managing structured content, the focus for the content management systems of the future may be less about humans managing individual content items, and more about managing the scaffolding, prompts, and governance rules for various AI elements and transformations at scale.
In the same way that databases now have more tools for federation, metadata, transformation and built-in logic, content systems will need to evolve capabilities in that direction. The early efforts in “content supply chain” will take on new importance as the understanding of where content was coming from or going to becomes more important.
Of course, this is just one example and suited to a large multi-national organization, but the potential for revolutionary AI elements can extend across numerous aspects of content management; content variations at scale, visual layout and imagery, iterative testing and improvement, etc. etc.
Ironically, as these technologies get better and better, you will hear vendors talk about “AI” less and less. For example, when Apple talks about the camera in their latest iPhone, they talk about improved colour rendition, performance in low-light, image stabilization and advanced lighting effects – all of which are incredibly advanced uses of AI to advance extensive computational photography, but AI does not make a single appearance in any of their marketing material.
We’ll know AI will have reached that “Plateau of Productivity” when we simply talk about the use cases that are enabled, and not the magical marketing pixie dust that will get us there.
This article originally appeared on LinkedIn.