As AI Makes Ad Production Easier, the Harder Problem Is Knowing What to Make

For years, digital advertising teams have dealt with a production problem. Video takes time to make, creators are difficult to source, and a campaign can require dozens of variations before a marketing team has enough intel to understand what’s actually working.

Artificial intelligence is now transforming the economics of that process. Generative tools can produce scripts, images, edits and other pieces of content at a fraction of the time and cost associated with traditional production. For growth marketers, though, cheaper production gives rise to a different question: when making another ad becomes easy, deciding which ad deserves to be made becomes more important. That is precisely where the emerging field of performance creative is starting to look less like a conventional advertising discipline and more like a technology problem.

Founded in Pasadena in 2018 by Francisco Cheng and Ken Chen, Narrative Ads has built its business around connecting creative production with performance data. The company has expanded from a creative execution agency into what it describes as an AI-enabled creator marketing platform with its own workflow software, production tools and creator network.

The shift reflects a broader change in how marketing organizations think about creative. Once a brand can generate hundreds of potential pieces of content, the scarce resource is no longer necessarily the ability to make another video but rather the ability to distinguish useful creative from noise.

When More Content Creates More Decisions

The economics of advertising production have been moving in this direction for some time. Social platforms have trained consumers to expect a constant stream of video, while marketers have become increasingly dependent on testing different hooks, formats and messages. A single campaign can produce enough creative variations to make manual evaluation impractical.

AI accelerates the process significantly. A marketer can ask a generative system for alternative hooks or variations on a concept in minutes. Editing and adaptation can happen faster as well. The resulting abundance is useful enough, but it shifts the nature of the workflow. A team producing 10 ads has one problem; a team producing 100 has another. The second team needs a way to determine which creative variables are worth testing, which creators are appropriate for a particular campaign, and which performance signals should influence the next round of production. Without those systems, additional content can simply mean additional work for the people responsible for reviewing it.

Narrative’s model was built around this very feedback loop. The company uses historical performance data, including click-through rates and elements of hook execution, when matching creators with campaigns. Its objective is to give marketing teams a way to make creator selection based on evidence rather than relying primarily on reach or aesthetic reputation.

The distinction is growing increasingly consequential for brands buying performance advertising. A creator with a large following may be valuable for awareness, but follower count alone does not tell a growth team whether that person can deliver an effective paid-social hook for a particular product.

Performance data offers a different way to evaluate the decision.

Turning Creative Into a Testable System

Narrative’s infrastructure has developed around that premise. The company has built a vetted creator network alongside a data architecture intended to track creative variables across campaigns. It also developed Narrative OS, its proprietary workflow suite, and Narrative AI, a collection of tools aimed at reducing friction in the production process. The company operates a 25,000-square-foot production facility in Pasadena with 11 real-world sets.

The combination points to an interesting development in advertising technology. AI can make individual production tasks faster, but the useful advantage may come from connecting those tasks to the information generated after an ad goes live.

Consider a basic creative test: a brand can produce several versions of a video, distribute them through a paid social campaign, and compare their results. The interesting information here isn’t necessarily that one version received a higher click-through rate but rather what the difference might reveal about the opening, the creator, the offer, the visual treatment or the way the product was introduced.

Once those observations feed into the next production cycle, advertising starts to resemble an iterative product-development process. Narrative’s production model operates on a weekly launch and iteration cadence.

The Creator Becomes a Data Point

The creator economy adds another layer to the equation. Brands increasingly have access to large pools of people capable of making native-looking social content. Finding someone who can appear naturally on camera is relatively straightforward compared to determining who’s likely to perform well for a particular campaign.

Narrative’s approach treats creator selection as a performance question. Historical metrics become part of the casting process, alongside the usual considerations of product fit and creative requirements. This has given rise to a different role for the creator. A person’s still responsible for the performance and personality that make a piece of social advertising work, but the surrounding system can evaluate their contribution using the same kinds of measurements applied to other campaign variables.

This also creates the possibility of compounding knowledge. If a company works across enough campaigns, it can begin to accumulate evidence about which creative approaches work for particular audiences and offers. Each campaign potentially adds another data point to the next casting or production decision. The challenge is maintaining useful signals as the volume of content increases.

AI Doesn’t Remove the Need for Creative Judgment

There’s an easy assumption that generative AI will eventually eliminate much of this complexity. If a system can generate thousands of ads, perhaps the marketing team simply needs to pick the winner. In practice, the selection problem can become harder.

A model can generate variations quickly, but someone still needs to determine what constitutes a meaningful test. Two ads can differ in dozens of ways, making it difficult to know which change produced the result. A campaign can also perform differently because of the audience, offer, placement or timing rather than the creative itself. For enterprise marketing teams, the technical challenge is therefore less about generating an unlimited supply of content than building a reliable process around it.

Narrative’s operating model’s built around accountability for metrics such as cost per acquisition (CPA) and return on advertising spend (ROAS), while agility around new AI technologies is another core value.

The approach puts creative closer to the financial mechanics of a campaign. A video’s valuable insofar as the performance data supports its continued use, adaptation or replacement. It also changes what marketers may expect from a creative partner. A production vendor can deliver the requested assets. A performance creative operation has a larger responsibility: determine what should be produced, get it into market, observe the response and use the result to inform what comes next.

A New Advertising Infrastructure

Narrative’s growth also illustrates how much infrastructure sits behind an apparently simple social video. The company’s network includes 136 specialized workers across the Philippines, Colombia and Nepal alongside U.S.-based talent. Its platform relationships include Meta Business Partner and Marketing Partner status dating to 2019, TikTok Marketing Partner status dating to 2021, and Snap Advanced Partner status. TikTok describes its marketing partners as companies evaluated for expertise on the platform, while Snapchat’s Advanced Partner program is intended for high-performing agencies and provides additional support and resources.

Those relationships are useful context, but they aren’t the central technology story. The more interesting question is what happens when production capacity, creator sourcing, campaign data and AI-assisted tools start operating as one system. For marketing departments, the answer could determine whether the flood of AI-generated content becomes an advantage or another operational headache.

The advertising industry has spent years trying to make creative production faster, and AI may accomplish much of that work on its own. The next competitive problem will be deciding what deserves to be produced in the first place, and building a feedback system capable of learning from the answer.

Narrative Ads has spent its business around that very problem. Whether its particular approach proves durable as generative AI reshapes the economics of production remains an open question, but the underlying challenge is likely to become all the more important as the cost of making another ad continues to fall.