What does AI video production cost, and can you scale video without wrecking quality? The short version: AI makes each frame cheaper, especially in pre-production, versioning and localisation, but it rarely makes the overall job cheaper, because clients respond to lower unit costs by asking for more. Agencies that keep pricing by the asset end up producing more for less. Those that price the idea, the system and the proof keep their margin.
Two stories this autumn show the squeeze from both ends. In Campaign Middle East, Assi Rahbani, managing director for MENA at TAG, said: "AI use in production has gone from optional to expected; clients ask about it directly in briefs now." And Campaign US reported that WPP Production had withdrawn internal guidance telling staff to "actively convince" clients to send all production work to WPP and avoid traditional triple bidding.
What does AI actually make cheaper in video production?
There is no honest universal price per finished minute of AI video, because the answer depends on how much of the job is the shoot and how much is everything around it. What you can pin down is where the savings sit, and Rahbani is specific. "Real value: pre-production," he told Campaign Middle East, adding that AI has "cut weeks out of the ideation phase". He names budgeting, scheduling, previz and localisation as the places to embed it.
Add the tasks agencies already resent: cut-downs, captions, manual resizing for every placement, versioning by market, and the performance report that follows. None of that is craft. All of it is time.
What AI does not make cheaper matters just as much. According to Rahbani, crew day rates, equipment and insurance in markets such as Saudi Arabia, the UAE, Qatar and Kuwait are climbing. And on culturally specific work, he warns that AI-generated content "still reads as synthetic to a local audience". His verdict: "a brilliant accelerant, but not yet a substitute for people who understand the region."
Which production stages should you automate first?
Map the job stage by stage and the pattern is clear: AI is strongest at the edges of production, weakest in the middle.
| Stage | What AI does well today | What stays human | Pricing implication |
|---|---|---|---|
| Ideation and scripting | Rapid routes, references, first-draft scripts | Choosing the idea worth making | Charge for the idea, not the hours |
| Previz and storyboards | Fast animatics and mood films for approval | Taste, casting, cultural judgement | Fewer rounds, faster sign-off |
| Budgeting and scheduling | Draft budgets, schedules, scenario plans | Negotiation, risk, contingency | Tighter quotes with visible contingency |
| The shoot | Limited: virtual sets and planning aids | Crew, talent, direction | Input costs still rising in many markets |
| Post and versioning | Resizing, cut-downs, captions, localisation drafts | Final edit and local sign-off | Price the system, not each asset |
| Reporting | Pulling results and drafting the narrative | Interpreting and recommending | Bundle into the retainer |
Why doesn't cheaper production mean cheaper budgets?
Economists call it price elasticity of demand: how much the quantity people buy changes when the price changes. When demand is elastic, a fall in price produces a bigger rise in volume. Video assets behave exactly like this. Make a cut-down cheap and clients do not buy the same number for less money. They buy many more.
Rahbani describes the result. Clients "are asking for the same, often more, with visibly tighter budgets", increasingly as integrated campaigns spanning film, social, activations and events "under one brief and one budget line". His summary: "Margin compression, plainly."
So if your fee is tied to units, AI hands the saving to the client and the extra workload to you. The way out is the insight behind this piece: value outshines price when connections thrive. The metric Rahbani uses to judge whether the industry is maturing is not cost per asset. It is "repeat business from the same client on integrated, cross-format work". Clients come back to the partner who makes the work land, not the one with the lowest rate card.
What does the WPP Production episode tell independents?
According to Campaign US, the withdrawn guidance said: "Our goal is to avoid traditional commercial triple bidding by proving the value of a centralised partnership." Staff were to offer three bids from within WPP instead. WPP's response: "Any suggestion that WPP Production undermines fair competition is fundamentally wrong." Elliot Liss, head of production at the IPA, said agencies "continue to work extensively with external production partners where that is the right solution."
Whatever you make of it, the economics are plain. When frames get cheaper, volume becomes the prize, and large groups want that volume in-house. Independents rarely win a volume war against a holding company's production arm. They can win on the idea, on speed, and on evidence that the work performs.
How do you scale video production without sacrificing quality?
- Split judgement from repetition. List every task on a typical video job and mark each one as idea, craft or repetition. Most teams find repetition is the biggest pile.
- Automate repetition first. Resizing, cut-downs, captions, localisation drafts, trafficking and performance reporting are the safest places to start, because errors are easy to spot.
- Use AI to widen the front end. Generate more routes and previz earlier, then spend human time choosing. Faster ideation is only valuable if someone kills the weak options.
- Keep cultural sign-off human. Before anything localised goes live, have someone from that market review it. Work that feels synthetic costs more in trust than it saves in budget.
- Track cost per usable asset. Not cost per asset produced. Versions nobody runs are waste, however cheap they were.
- Report outcomes, not output. Show the client which assets drove results and what you changed because of it. That is what earns the repeat brief.
How should agencies price AI-assisted video?
Stop selling frames. Sell three things separately: the idea, the production system and the proof.
- The idea carries a fee that does not shrink because rendering got faster. Judgement is the scarce input.
- The production system, meaning templates, automation and the versioning workflow, is priced as a capacity retainer rather than per asset, so volume can rise without renegotiating every cut-down.
- The proof, meaning reporting and optimisation, is built in, because it is what brings the client back.
Imagine a 20-person agency asked for 60 social versions of a hero film. Priced per asset, every efficiency it finds becomes a discount. Priced as a system with a capacity ceiling, efficiency becomes margin, and the agency can afford more senior time on the film itself.
This is where the Defy platform fits. Defy Agents run agentic workflows for repetitive delivery such as reporting, monitoring and resizing, so people stay on the work that wins. SAMY, the global influencer agency, used Defy Agents to cut performance reporting from three days of work to one.
