Tools are not (ultimately) important. Something else is.

Ever since generative AI became a thing in content production – all of 3,5 years ago – what we’ve predominantly talked about (if we’ve been interested in applying gen AI in our work) has been the tools. Should we use Midjourney or Dall-E for imaging? Is ChatGPT suitable for writing treatments or offer feedback on dialogue options? Is is Sora or Runway or something else we should turn to to spice up our videos? 

This is all natural, especially looking at the speed with which the world of AI is moving – it always feels like the next big thing is just around the corner and will offer you something that no previous model has been able to. But while these programs and models and tools matter, and while they lure you in with bells and whistles and promises they are ultimately a distant second to the planning and structural thinking that goes into deciding how a project should be designed, built and finally experienced. 

So while AI does offer an ever-expanding and evolving toolkit, what we need is not so much that as a better architecture. We should strive to build a framework to decide the role of AI in our creative processes, how we envision AI interacting with audiences and with creativity, and find out where AI actually should be placed in the creative process. 

The five layers

I’m proposing a framework that consists of five layers, clearly differing from each other while at the same time clearly building on each other. This is not intended to be a strict guide but rather something that can help me – and hopefully you – view AI:s role in a fruitful light. I’m all ears if you have suggestions, comments, agree or disagree!

First layer – the core of the story

I’ve had many discussions with creatives around the use of AI. The majority are very averse to letting AI have a say in deciding the core of the story. And I agree – this layer should be AI-free. Why do I want to tell this story? What is this project about? Why does it matter? These are all questions that AI should not be answering for us. 

In a previous post, years back, I included a “creator’s checklist” in a questionnaire I built. This layer is where we answer some of those fundamental questions regarding our project, our story: What are my goals? What world does the core story reside in? What kind of impact can I hope for and strive for?

And yes, this is the layer where we make some crucial decisions regarding representation, diversity, inclusion, cultural sensitivity and so on. This is all best done before we implement any AI whatsoever. 

Second layer – the ecosystem

According to the transmedia storytelling principles I’ve always felt are sound and extremely useful to apply to any project I’m working on, to a higher or a lesser degree, this layer would be where the transmedia design happens. Using the core story and the storyworld as foundation, we decide which platforms to use, what stories get told where, what entry points the audience are given, how different stories interact and support each other. This is an area where AI can be very useful, as a never-tiring partner that can offer suggestions, comments, help identify gaps or potential pitfalls, and draw up models for how stories can connect from one platform to another. It can help you maintain your story bible by updating it whenever new decisions or inclusions have been made, ensuring that the bible is anything but a static document. 

Remember, it’s still you in the driving seat, no matter how many suggestions AI throws your way. 

Third layer – production 

This is where AI and film (and media overall) works the most together in the real world at the moment. I can use AI to help me visualise ideas, to restore archives, to colorise images and films, to translate and transcribe, to assemble a first rough-cut, to compose or change music, to use effects… all of these things and many more are represented in this layer. And if you, like me, work a lot with documentaries, it has the possibility to be a god-send when it comes to analysing terabytes of interview footage, find themes, and flag suggestions for connections that humans easily can miss. 

AI in this layer has one specific line drawn. It does not make any creative decisions of its own; it’s beholden to the creative decisions made in the first two layers. 

Fourth layer – AI and the audience 

As I was alluding to before, I feel AI can finally allow us to succeed with ambitions that were simply impossible (or very nearly so) before. Building on the three previous layers, this is where AI has the possibility to offer audiences responsive, personalised storytelling. An AI system firmly grounded in the story core and the design decisions taken in previous steps and with a firmament consisting of a thorough story bible that can, without missing a beat, adapt storytelling and experiences without breaking anything and offer them to the audience. Furthermore, an AI system that can accept audience interactions and contributions and place them in their right place in an evolving story system. 

This might sound far-fetched, but with AI being the worst it’ll ever be right now, and for this particular area I believe it’ll already now be very good, I have high hope that we will reach systems like this sooner rather than later. If the audience, as Jeff Gomez talks about in his seminal posts about the Collective Journey, becomes a protagonist, we must have a system that is capable of taking on inputs from a number of sources, while still maintaining the integrity of the core narrative. 

Still, this is also the layer with the most risks. AI can for instance, and models have been known for reflecting biases from its training data. What sets out as personalisation might become manipulation of some sort. Whatever data the audience provides us with can be exploited in many ways. To combat these and other risks, they must be taken into consideration in the first layer and serve as the foundation for this fourth layer.

Fifth layer – feedback

The one thing we want (or at least should want) is feedback from our target audiences (and others as well, of course). How did we succeed? What can we do better? What would make audiences return, again and again? While focus groups and data analysis are both great tools, AI can jump in here too. It can help to, quickly and efficiently, analyse the engagement of the audience. Where do they engage the most? What content is valuable? When do they tune out?

With the help of AI, this doesn’t just become numbers and statistics, open for interpretation. It can help put the data into the right context and find solutions and possibilities. In that same questionnaire I launched years ago I talk about different aspects to take into consideration with creating something that is intended to resonate with the audience and invite engagement. I talked about intensity and polarisation, longevity, and what I called the flash-versus-slow-burn dynamic. AI can measure these dimensions in real time, across platforms, at a scale very difficult for a human to manage.

Circling back to Jeff Gomez and his notion that having someone really genuinely listening to you is one of the core aspects of feeling seen, heard and respected as a fan… Well now AI can let us listen at a scale previously impossible. We can now listen to thousands – no, millions – of audience members at the same time, while retaining their identity as participants. 

 If genuine listening is at the core of any fandom, as Gomez argues, then AI can help creators listen to millions of audience members simultaneously – not as data points, but as participants whose engagement patterns reveal what the story means to them.