“The Goal Is Not to Generate More, It Is to Choose Better”: Simon Percy on Productions Embracing AI
Simon Percy has been working in the industry for over two decades. Starting his career editing music videos, honing storytelling skills through creative visuals which sparked his passion into visual effects and animation.
In 2000, Simon started Darkside Animation where he ran the day to day operations of the studio. There he refined his craft, building pipelines and running artists across an array of premium film and television projects for Universal, MGM, Discovery, ITV and the BBC.
Never losing touch with the team and sight of the tools and processes that enable the work to be created, Simon would regularly be found working on the productions he was running, understanding that an integral part of the process is knowing what is possible and how to achieve it.
Simon spoke with LBB about the quality improvements that have come from VFX embracing AI, the ethical guardrails put in place, and tackling the ‘creative drift’.
LBB> What is the most impactful, specific task or process that AI has permanently changed in your current role, and how does it drive qualitative improvements (not just efficiency) to your final output?
Simon> The permanent change is front-end look development and decision-making. We can now explore a far wider set of creative options early, then lock a direction with much more confidence before we commit serious production time.
The quality improvement is real because it reduces guesswork. We can test mood, lighting behaviour, texture response, production design choices, and even continuity across a sequence, quickly and repeatedly. That means fewer late stage compromises, fewer ‘good enough’ fixes, and a cleaner path to a cohesive visual identity.
In broadcast and commercial, where timelines are tight and feedback is constant, the ability to pressure-test ideas early is often the difference between something that feels premium and something that feels merely finished.
LBB> AI output is only as good as the input. What is the single most valuable lesson you've learned about prompt engineering or interaction design to get truly distinctive and high-quality creative work from an AI model?
Simon> Stop thinking in prompts and start thinking in constraints. The best results come when you define what must not change as clearly as what you want to see. For us, that means anchoring the work with strong inputs (owned or licensed reference, clear art direction, camera language, and a defined finishing target), then iterating within tight boundaries. If you leave the model too much freedom, you get generic ‘average of the internet’ aesthetics. If you treat it like a junior artist who needs very specific direction, you get distinctive outcomes that still feel authored.
LBB> The ethics of using AI are centered around data sourcing and IP. What is the most critical ethical guardrail or internal policy your organisation has implemented to ensure your use of AI to generate or assist content is responsible and legally sound?
Simon> The critical guardrail is provenance-first inputs, full stop. If we cannot evidence rights to the inputs, we do not use them. Operationally, that means: we prioritise client-supplied assets, our own photography and scans, or properly licensed materials. We keep work in controlled environments, we do not train external models on client data, and we maintain an audit trail of key creative decisions, versions, and derived assets. We are model-agnostic, but we are not policy agnostic. The process has to be defensible to a client, a commissioner, and a legal team.
LBB> What is the biggest challenge in collaborating with AI as a creative professional? Specifically, how do you work to ensure that the final work maintains a demonstrable sense of authenticity, human intuition, or your signature creative identity, and how do you personally avoid getting sucked into dependence, ‘work slop’ mode or switching your critical brain off?
Simon> The biggest challenge is creative drift. AI can pull you towards the easiest, most statistically likely answer, which is where ‘work slop’ comes from. We counter that by keeping authorship explicit. We define the visual rules early (composition, lens language, lighting intent, texture character, pacing), then we enforce those rules with the same discipline we would in any traditional pipeline.
AI is a tool in the middle, not the decider at the end. Personally, the safeguard is simple: We do not accept outputs at face value. Everything gets interrogated for taste, story, continuity, and believability. If a result is ‘impressive’ but not purposeful, it is noise. The goal is not to generate more, it is to choose better.
LBB> Have you found that the integration of AI tools creates new points of friction or collaboration challenges with other creative or production partners? If so, how are you overcoming them?
Simon> Yes, mainly around trust, handover clarity, and expectations. Some partners assume AI means instant results, while others assume it means legal risk or creative compromise. We overcome that by being transparent about processes rather than talking about tools. We explain where AI sits in the pipeline, what is being used as input, how we control outputs, and what the approval gates are.
On shared projects, we also define what is deliverable at each stage, so nobody mistakes exploratory work for final. That clarity reduces friction, protects relationships, and avoids scope creep disguised as ‘just try one more’.
LBB> As AI rapidly changes workflows, the skillsets required are evolving. What is one new skill or capability related to AI usage that you believe is essential for people in your discipline to learn in the next year?
Simon> Creative evaluation and control. Not generating, judging. In practical terms, artists need to get good at: defining constraints, building repeatable workflows, spotting artefacts and continuity breaks instantly, and understanding provenance and licensing enough to work safely. The value is shifting from ‘who can press the buttons’ to ‘who can consistently steer the work to a premium, defensible finish’.
LBB> There’s some evidence that the productivity gains of AI have been overstated at an organisational level. How are you making sure that you’re using AI in a genuinely effective and impactful way, and are there any areas that you’ve discovered that AI tools can have a detrimental impact on productivity/effectiveness?
Simon> The productivity gains are real, but only when you measure the whole system, not just the generation step. AI can easily increase churn: more options, more review, more indecision, more versioning chaos.
We keep it effective by tying AI usage to specific production outcomes: fewer rounds, earlier sign-off, stronger consistency, or reduced rebuild time. If it does not move one of those needles, it is a distraction.
Where it can hurt productivity is unstructured exploration, unclear approvals, and poor asset discipline. If you let it run wild, it creates a mess of near-duplicates and subjective debate. So we time-box exploration, we lock decisions, and we treat outputs like any other asset with naming, version control, and QC.
LBB> Do you think AI is currently creating entirely new forms of art or media that weren't possible before?
Simon> Yes, particularly in scalable, authored variation. Not ‘infinite content’, but coherent systems that can produce multiple campaign outputs without losing identity.
In advertising, you can now build a brand world with consistent art direction, then generate controlled variations for different markets, formats, seasonal moments, and performance edits, far faster than traditional reshoots or full rebuilds. The new medium is not the single film, it is the responsive campaign system, as long as it is guided by real creative direction and proper guardrails.
LBB> Thinking about your own role and discipline, what kind of specific, measurable impact do you think AI will have in the medium-term (3-5 years)? To what extent will it change the way people in your role work?
Simon> In 3 – 5 years, the measurable shift will be in iteration speed and the ratio of time spent on manual labour versus creative supervision. You will see smaller teams delivering higher shot counts and more variants, with a bigger share of time going into look definition, review, and finishing polish.
For my role as a studio owner, it will increase the importance of process, compliance, and repeatability. Clients will not just buy visuals, they will buy confidence: that the work is premium, consistent, and legally clean. The teams that win will be the ones who can combine taste with a robust, policy-led pipeline.
