Here’s your complete Ai strategy: Implement, normalise and get back to comfortable margins
Ai is vital, yes, but the caveats are outrageous. Here’s what you need to do to stay sane:
1. Do shoehorn in as much gen Ai as possible → don’t rely on it for delivery
When it works, great! It is near instant and it is near free. It has changed what's possible. But it’s practically useless on its own. It is not a delivery pipeline. It's a black box. You get sliders & prompts → a few simple controls. Then you must take it or leave it — it's infuriating. The main players, MJ, Runway, Veo, and the others in the 'instant video' crowd, they could all be making ComfyUI style interfaces that give you control, but they don't. Complexity kills the consumer business model. It's pegged back & can't hit your brief. It’s not broadcast quality, or commercially cleared for use. Use it for
• Beautiful storyboards → Go wild in Creative look dev
• Use it for generic backgrounds, plates & filler shots
• Best of all, it’s amazing for ‘bits’, a whiff, a splodge, or a hint..
It will help with all the hero moments & a complete ad isn’t totally out of the question, just very rare.
No one using instant video alone will stay in a business that delivers on brief. You could just press those buttons yourself
2. ComfyUI is the future: the perfect complimentary workflow - but it is not widely implemented yet
Ultimately, and very soon, you will construct exactly the video you want, bit by bit. ComfyUI is leading with tools that modify the image at the point of creation — You get creative control tailored to your brief.
Comfy isn’t in the mainstream yet. Nearly! There’s new technology to learn. Industry training is only just getting underway. Pro kit and expert engineering are essential. It’s not a good bet to pull in any ComfyUI Freelancer you find, you’d need to know it’s been rolled out responsibly to make sure it’s a stable solution for your production.
I’m currently deep into my 2nd stage training, exploring best practice for ComfyUI in Production.
Training from ActionVFX
I started ComfyUi foundation training early. I can brief clients & production on how it works and when to use it. I am on track to offer ComfyUI in production by late Oct 2025, but I’m not going to jump before I’m ready.
3. Don’t rush in → even if you did nothing, your supplier is implementing Ai anyway
The panic is over. We’ve seen what Ai can do. What we need to do is clear.
Every VFX app you’ve heard of is already jam packed with Ai. Flame runs models through its inference node. Mocha makes garbage mattes in a click. Resolve's neural engine gives clean roto & if you train Nuke's Copy Cat it will do your whole shot. This is the current hybrid and it is working now. VFX kit & know-how are still required. None of it is instant, but it's getting faster every day. It's not free, but it doesn't break the bank. It is stable, flexible & delivers broadcast-ready on brief.
4. The last thing is asset preparation, integration & assembly - the new delivery essentials
Whilst more beauty and effects go to Ai, more integration & assembly is coming to Flame. There is still no Ai asset that comes out ready to use and there is no way to deliver without experience. You’ve probably heard more about Ai of late, but Flame continues to dominate commercials delivery in 2025 and has doubled down by consolidating its toolset as a delivery hub. That’s where it shines!