Balancing AI and Human Craft 

Covert founder and ECD Simon Dewey unpacks how post houses can inject AI into traditional workflows, speeding up labour-intensive tasks while protecting human craft, writes LBB’s Alex Reeves

​The post-production industry is facing a new, deeply frustrating public relations crisis. Whenever a piece of heavy VFX or CG work drops, the immediate assumption from the public – and increasingly, from agencies – is that it was simply conjured up by artificial intelligence.

It is a reflex that dangerously devalues the painstaking human craft required to make great work. Simon Dewey, founder and executive creative director at Covert, knows this frustration all too well. He argues that this assumption cheapens the final image, ignoring the specialists working away behind the screens.

To combat the online trolls claiming AI is stealing jobs, post houses are now having to work overtime just to prove their human labour. “There’s a near-term solution which is suddenly VFX breakdowns are worth so much more than the piece,” Simon explains. “Everything you put out, you almost need to show the hardware render along with the final piece.”

The irony is that integrating AI at the highest level of post production requires equal, if not more, effort than traditional methods. Pointing to a recent NHS project featuring an AI-assisted hamster, Simon notes that reaching a premium, photorealistic finish is intensely demanding work. Far from rendering artists obsolete, the reality of wrangling AI into a professional pipeline has actually created jobs. But changing the narrative requires an enormous educational effort to show clients exactly how much traditional craft goes into the ‘AI’ process.

But this education cannot be a one-size-fits-all data dump. Simon warns against overwhelming uninitiated clients with a rigid, highly technical breakdown of the studio’s workflow. “We actually need to break down how we educate to several tiers,” he explains. If a post house simply presents its standard 25-step internal process to a brand that assumes AI is a one-click solution, the message loses all nuance. But also going into too much detail is baffling. “That’ll be so heavy for people that [are] uninitiated with working in this way,” Simon notes. Instead of getting bogged down in hours of technical explanation to prove why the work isn’t “just AI,” communication must be tailored to the client’s baseline understanding – proving the value of the human craft without drowning them in jargon.

As Covert’s team often experience it, the biggest misconception agencies harbour isn’t about what AI can do, but rather what it finishes. While the generative capabilities of modern tools are undeniably vast, relying on them to spit out a final, broadcast-ready asset is a mistake.

“I almost split the question into more what it can’t do,” Simon explains. He notes that while AI can theoretically tackle almost any task – from generating characters to replacing objects – “it can’t achieve a fully realistic result without lots of craft.”

Instead of a magical finish button, its true power lies in its potential to upend existing heavy-duty processes, such as virtual production. Rather than relying entirely on massive LED volumes, directors can focus purely on capturing a stellar on-set performance, leaving the world-building to post.

As Simon points out, “AI is very quickly going to become the new world of creating, whereby performance is still very important… But in terms of being able to shoot a scene and then change the environment, relight the environment, VP is already redundant.”

However, this generative flexibility is a double-edged sword. Without traditional post-production techniques stepping in to refine and ground the output, the raw generation remains unusable for premium brands.

To bridge this gap between a broad-strokes generation and a photorealistic final shot, studios must rely heavily on established, multi-step pipelines. The central tension, Simon argues, comes down to quality control.

Generative tools offer incredible speed and flexibility, allowing teams to build assets and block out environments efficiently. But brand clients inevitably demand absolute precision – the ability to tweak a specific lighting reflection or adjust a subtle colour grade.

“‘There’s a new concept we need to think around which is controlling the quality,” Simon notes. While he acknowledges that post houses can build elements using AI “very quickly, efficiently, easily”, he stresses that the moment a client requires granular adjustments, “that’s where we have to work so much harder in traditional terms.”

This is why foundational craft remains non-negotiable. An AI tool might only handle a small fraction of the total steps required to finish a shot. Tying generated elements together seamlessly requires a deep, traditional knowledge base of how light, movement, and texture should actually behave. As Simon puts it, it is only “through having a knowledge base of how things should work, how they look” that post-production professionals can actually deliver premium work.

When looking at where AI thrives in a high-end pipeline, the use cases fall into two camps: the functional and the foundational.

For labour-intensive prep work like rotoscoping, AI is a seamless addition. “It will cover several use cases because it’s non-creative and it’s functional,” Simon explains.

Creatively, it is heavily utilised to generate references, aligning teams before the real work begins. Yet, its most exciting application at Covert is acting as a hyper-fast render engine. Simon points to an early test where the team built a basic CG parrot to retain absolute control over its animation, but used an early AI model – the first version of Sora – to render the photorealistic final details. “The output was so strong,” he notes. “This is where it’s going.” By layering AI rendering over a traditional CG build, studios can successfully marry the granular control of human craft with immense generative speed.

To truly understand AI’s place in high-end post production, the industry needs a fundamental perspective shift: raw AI output is unfinished work. Much like untextured CG elements, an AI-generated asset is merely a starting point that requires traditional compositing to feel grounded and realistic.

Simon likens the current generative boom to the advent of digital photography. Just because powerful tools have suddenly become widely accessible doesn’t mean the user instantly becomes a master craftsperson. The application is highly nuanced.

In practice, a premium post house doesn’t simply prompt a video and drop it straight onto a timeline. “If we are creating with AI, we don’t just create with AI, we’ll break down a shot to all of its components,” Simon explains. Instead of generating a whole scene, artists generate individual elements or specific objects. From there, they “add little imperfections, combine it and develop it.”

As soon as it is viewed through that lens, AI is no different to any other raw material. It demands the meticulous, human hand to actually finish the job.

Ultimately, the greatest threat AI poses to the creative industries is not job theft, but the dilution of specialisms. When agencies or directors oversimplify the technology into a mere prompting exercise, they risk severely watering down the quality of their campaigns.

“Craft is absolutely everything,” Simon warns. He argues that an ignorance of traditional post production is precisely “why AI is not being taken seriously” in so much of the broader media landscape right now.

While a director might be able to string together an AI-generated sequence, doing so distracts them from their primary role: steering the creative vision. For the industry to maintain its premium standards, execution must remain in the hands of dedicated experts who actually understand the nuances of the image. As Simon concludes, “Yes, you could do a bit of AI, but it makes far more sense when you’ve got teams completely dedicated to that to really lean into those directions. It’s important that we don’t spam and start to release these things because it just means a watering down of quality.”