Think of AI as doing the laundry so we can create better art   

AI has pressed fast forward on agency production models in recent years, accelerating pre- and post-production tasks, unlocking new visual possibilities and, in some cases, reducing costs. But the rapid rise of generative AI has not come without its bloopers: IP concerns, ethical quandaries and a public that is increasingly sharp-eyed and sharp-tongued when it comes to spotting and criticising AI-generated work. 

How much has AI already become part of your day-to-day production workflow - and what kind of impact are you actually seeing?  

It’s been a mad past year for AI, almost impossible to keep up with the pace of change.  We’ve been encouraging teams to explore tools and are seeing impact across the business. We are currently in a test-and-learn stage across many aspects, and working on governance before AI is fully embedded across the business. It’s definitely a learning curve, and the more time spent getting to know the tools, the better the outcome proves to be.

We're seeing impacts mainly in enhanced proof of concepts, guide VO's, translations and subtitling, and in efficiencies in small-scale post-production tasks. In a time when GenAI is grabbing all the headlines, it's easy to forget the other AI-powered tools in a post-production workflow, many of which use AI or machine learning to add efficiency to everyday tasks.

These tools don’t replace human craft and they’re not always 100% accurate in their output, but when you need to perform a repetitive task like rotoscoping or creating some temporary VFX for an offline edit, these tools can give you an edge in enhanced quality or efficiency; taking a task that previously may have taken hours (or days) and boiling it down to just a few minutes.  

What kinds of tools or techniques have proved genuinely useful, rather than just experimental?  

We've been playing around with a variety; Kling, MidJourney, ChatGPT, Gemini, ElevenLabs, Nano Banana and Higgsfield. In production, image-based tools (e.g., MidJourney) for mock-ups/scamps are proving very useful, as are ElevenLabs for guide VOs. As I mentioned above, in post-production, the useful everyday tools like Nuke’s CopyCat, AE’s Rotobrush, Resolve’s Magic Mask, and AI upscaling from applications like Topaz. There’s certainly been a lot of experimentation to get to this stage, and like any transformational change in a business, it's an iterative process with plenty of failure along the way.      

In practical terms, where does AI make the biggest difference - speed, cost, creative iteration or something else?  

Currently, we’re finding it very useful for creative iteration and pre-vis stages. Where we perhaps used to rely on found imagery, we can now create much more specific scamps to help us explain and sell in ideas, allowing clients to further minimise risks. This is not to replace the final output, but rather to better visualise in the early stages and give a wealth of options at a fast pace. I hope that, once our governance is more embedded and AI is fully immersed into our workflows, we’ll start to see cost implications too.    

Can you share a recent example where AI helped your team deliver something faster, smarter or better than before?   

We’ve recently been exploring multiple creative routes for research, and whilst AI hasn’t been the complete solution, we’ve used it to enhance the process while still employing storyboard artists. This has included playing around with multiple AI VOs to deliver the research. Not every voice used is AI, but it’s allowed us to pepper in more characters and, overall, create more varied boardmatics at a lower fee. This has also allowed the creative development stage to be more comprehensive, as we’ve been able to try more voice options for each character to get to the funniest and most impactful outcome.  

What are the moments where you’ve found AI still can’t match human craft or instinct?  

We are excited to embrace AI across various points of the work stream, but we are strongly of the opinion that

AI is never a replacement for human craft and instinct.

It is a beneficial, efficient way to facilitate vision and enhance how we get to our objectives, but the craft and idea always come from the human in the loop: human judgment and nuance ensure the highest quality.   

If you take the skill of a DOP as an example, AI can suggest camera angles (like Higgsfield’s recent update) and lighting set-ups, but it can’t replicate the experience, meticulousness and reactivity of a skilled DOP. Similarly, AI can pull together a rough edit, but only a human eye can understand the difference that moving a cut point by just a couple of frames can make. 

How do you balance automation with the need for creative control and quality - especially when clients expect both innovation and polish?   

It all starts with questioning “would AI make this easier/quicker/better or would AI be a hindrance?” You can’t just jump to AI and expect it to solve everything. If AI is going to be a useful aid or an enhancer, you then need to identify where in the process, and from then on it’s about combining tools with thorough human quality control. Yes, clients expect innovation, but not at the cost of good creative and high-quality output that is going to get them results. Work that gets the best results has the surprise and originality that demands human thought.    

Has AI changed how you scope, schedule or cost productions and are timelines or expectations shifting as a result?         

Are there parts of production you think AI will transform next?             

I think the focus will be on pre- and post-production workflows rather than core creative decision-making.   

That being said, for creative ideation, I think the “AI mistakes” might start becoming part of the process. I think we’ll start seeing a lot of “I didn’t mean it to take me here, but it’s moved the creative on in xyz ways”.   

But within pre-production, things like concept visualisation, exploring more ideas quickly and helping clients understand ideas more confidently before committing budget, are all seeing impacts. I think shortlisting on cast or locations could be tapped into more, whilst keeping final decision-making with humans, of course.   

In post, machine learning has been slowly helping processes for a long time on a small scale, but I can see versioning and localisation being the most impacted moving forwards (which is already happening at speed!). Also, the post-production tools I mentioned above are where AI is singing.

If you’re looking into these new technologies you’ve probably come across the idea that

“I want AI to do my laundry and dishes so I can do my art and writing, not for AI to do my art and writing so that I can do my laundry and dishes” and that’s the good thing about all of the applications mentioned - they speed up the boring stuff (the metaphorical laundry) so you can be more creative.   

There’s no doubt that GenAI has a place in our workflows and is certainly here to stay, but when the dust settles on the hype, AI-enabled smart tools could be the key to unlocking a more creative future (with less laundry).    

What does this shift mean for people - how are you upskilling or re-framing roles within your production team?              

We are filmmakers, and we love craft. AI is a new tool to add to our skillset, but not replace our teams!

The output of Gen AI can only be as good as the information it’s given, and this is where the upskilling comes in. Growing and learning has always been about adapting to current trends, new tech, new connections - and AI is just another iteration of this.   

There’s a shift in where time is spent; it’s a great thing that, for example, production designers can accelerate mock-ups and spend more time on sourcing and building the real-world sets. Or it’s about front-loading work, because ensuring your prompts work as hard as possible is akin to writing a really tight brief. The storyboard process is a good example, where prompts need to be thoroughly considered to facilitate the output required. There’s learning to be done to translate production language into inputs that AI can actually understand.  

We’re all learning, and the models are changing at an enormous pace. Anyone who says they know with any certainty the direction that AI is heading is fibbing! But as hungry filmmakers, we’ll continue to adapt and innovate in order to make the best work possible; with AI being a key player.   

Looking ahead, what excites you most about AI’s potential in production - and what keeps you cautious?             

The potential is huge - and there’ll be so many advances that seem unfathomable right now that’ll soon make this write-up seem very out of date, I’m sure!   

I hope we get to a point where AI helps us shape considered and efficient production workflows without removing creativity and innovation. The ideal is that AI helps humans enhance their work, and allows us to do more with less, always staying true to our ultimate goal of connecting with our audiences.  

I think decision/risk support is an interesting part. Will there be ways in which producers can quickly stress-test different decisions and ambitions before locking into a route?   

You’ve got to balance cautiousness with being curious.

Producers come in all shapes and sizes of risk aversion: I’m on the slightly safer side and hope that everyone is always fact checking, sense checking and ultimately giving every decision a “human test” before doing anything rash with AI!