When AI Makes Content Cheap, More Assets Are Not the Answer

If generating a video cut, changing an aspect ratio, or localizing a voiceover keeps getting cheaper, should a brand simply convert the savings into more assets?

Usually, no.

On September 30, 2026, the World Federation of Advertisers reported that 96% of the major brands in its sample were using generative or agentic AI. Eighty percent were using AI for content creation, while 74% were using it for ideation and production. On the same date, Bain published a benchmark study in which only 6% of marketing organizations said AI was delivering significant performance impact.

Those figures are not a subtraction exercise. The studies used different samples, questions, and definitions of adoption and impact. They do not prove a 90-point failure rate. Together, however, they expose a useful procurement truth: access to AI is becoming common; the operating system that converts it into better decisions and measurable learning is not.

For global brands connected to China’s supply chain, that distinction matters. Fast access to samples, factories, and production partners can make asset creation very efficient. But efficiency can also multiply an unclear product claim across six markets, five formats, and dozens of paid-media variants before anyone notices that the original decision was wrong.

The content budget therefore has to answer more than “How many videos can we make?” It must also answer “Who decides what deserves to be made, what evidence makes the claim credible, and what result tells us to stop?”

A lower cost per asset can hide a higher cost of being wrong

In traditional production, expensive execution forced teams to make difficult choices before the shoot. When a campaign required sets, talent, a crew, and a substantial post-production schedule, the product story and approval path usually had to be settled early.

Generative AI removes many of those production constraints. That is valuable—but it can also turn an unresolved brief into 40 polished versions.

The true cost of a content program has at least four components:

1. Decision cost: customer research, product truth, audience priority, and the creative hypothesis;
2. Production cost: live action, AI generation, editing, sound, localization, and adaptation;
3. Distribution and learning cost: media, test design, reporting, and review;
4. Wrong-direction cost: rework, wasted impressions, missed timing, and damage to product credibility.

AI reduces the second component first. If decision quality and learning do not improve with it, the fourth component can scale faster than the content library.

Adobe’s 2026 AI and Digital Trends report shows the same tension from another angle. In its survey of 3,000 executives and practitioners, 76% reported improvements in the volume and speed of content ideation and production. Yet only 44% had implemented a measurement framework for generative AI, only 31% had one for agentic AI, and 47% had neither framework or were unsure whether one existed.

More output is evidence of capacity. It is not yet evidence of better marketing.

Automate the repeatable work; assign the accountable decisions

“Keep humans in the loop” is too vague to guide a budget. A better division of labor starts with accountability.

IAB’s March 2026 guide to agentic AI in video advertising describes media work moving toward AI-driven execution with human oversight and guardrails. The point is not to keep a person clicking through every step. It is to keep responsibility attached to the decisions that can change the brand, the claim, or the budget.

The ONCE Creative Decision Brief

Before a brand scales production, ONCE recommends a one-page Creative Decision Brief. This is not another machine-readable operating brief for an agent. It is a human-owned record of why a piece of content should exist in the first place.

The goal is not to predict the winning idea perfectly. The brief makes the learning legible. Without it, ten variants are often ten cosmetic changes. With it, variants can represent different hypotheses, and the result of one cycle can inform the next.

That distinction is increasingly important as strategic judgment becomes scarcer. In a separate September 22 study, WFA and mediasense found that 72% of 166 industry respondents identified strategic thinking as marketing’s biggest talent shortage, ahead of data strategy and execution at 46% and insight development at 36%. The sample included representatives of 65 brands. The result does not prove that every marketing team lacks strategy, but it helps explain why cheap execution has not made good prioritization cheap.

Three budget pools, adjusted by uncertainty

Brands do not need a universal 20/60/20 formula. A new regulated product, a proven household SKU, and a rapidly iterated social offer carry different risks. The more durable approach is to protect three budget pools and change their relative weight as uncertainty changes.

Decision budget

This covers customer-question research, competitive context, product-fact verification, creative hypotheses, prototypes, and approval rules.

Increase it when the product is new, the market is unfamiliar, the claim is easy to misunderstand, the category carries regulatory or reputational risk, or a reshoot would be expensive.

Production budget

This covers live action, AI generation, editing, sound, versioning, localization, and delivery management.

Increase it when the product truth is stable, the core proposition has evidence behind it, and the main task is adapting a proven creative system across markets, channels, or formats.

Learning budget

This covers test-cell design, media, reporting, customer-service and comment analysis, review sessions, and changes to the next production cycle.

Protect it when the brand runs always-on media and receives fast feedback. Bain’s September 30 analysis found that leading marketing organizations were 8.5 times more likely to run 100 or more experiments per month. The practical lesson is not to chase an arbitrary experiment count. It is to build a system in which production leaves behind a decision the next cycle can use.

The allocation rule can be summarized simply:

• The higher the direction uncertainty, the more you invest in decisions.
• The more stable the facts and reusable the production base, the more you can invest in scale.
• The faster the feedback and the more continuous the media program, the more learning budget must be protected.

A better way to compare production partners

Cost per video still matters, but it should not be the first procurement question. Ask potential partners:

1. What commercial decision will you help us clarify before production begins?
2. Which product claims require live-action or otherwise verifiable evidence?
3. Are the variants testing different hypotheses, or only changing surface features?
4. Which checkpoints require a named brand, strategy, or production owner to approve the choice?
5. When performance data returns, what exactly changes in the next batch?

A supplier that can only describe equipment, production days, and deliverable counts is selling capacity. A partner that can also define the hypothesis, proof threshold, decision owner, and learning handoff is selling decision quality alongside production.

This is the more valuable direction for ONCE as well. The long-term service is not simply “more videos, faster.” It connects a Creative Decision Layer, a Production System, and a Learning Ledger: decide what deserves to be made, expand it through live action and AI, then turn performance evidence into the next reusable judgment.

The commercial model can evolve with that structure. Our earlier analysis, AI Agents Are Becoming Metered Labor: How Brands Should Buy Creative Operations, explains why seats and raw output counts are weak proxies for agentic work. The same principle applies here: brands should buy accountable decisions and successful work units, not an unlimited pile of undifferentiated files.

AI will keep making “one more version” cheaper. The budget worth protecting is the one that decides which version deserves to exist, what truth it must show, and when the team should stop producing.

Sources and interpretation note

• World Federation of Advertisers, State of AI in Marketing 2026 press release, September 30, 2026; 54 senior respondents from 46 major brand owners.
• Bain & Company, AI in Marketing: How Leaders Achieve Double the Revenue Impact, September 30, 2026; 2026 Leaders and Laggards Survey, n=1,397.
• WFA and mediasense, Humans, Marketing and Machines press release, September 22, 2026; 166 industry respondents.
• Adobe, 2026 AI and Digital Trends Report; 3,000 executives and practitioners plus a separate survey of 4,000 customers.
• IAB, AI-Powered Video Outcomes: Agentic AI, March 30, 2026.

The studies use different populations, geographies, questions, and definitions. This article uses them to identify a shared decision pattern; it does not merge their percentages into a single industry failure rate or treat correlation as causation.