The ROI of AI content: how to really measure it and when it pays off.

The ROI of AI content isn't measured only by the savings on a single video. It's measured across three levers at once: how much each piece of content costs, how quickly it reaches the market, and how many variants you can test to find the one that sells best. Anyone who looks only at the price of the commercial misses the most interesting part of the return.

The major consulting firms confirm it. According to McKinsey, generative AI can increase marketing productivity by a value of between 5% and 15% of total marketing spend. BCG reports clients achieving up to 40% savings on production costs and ten times more content.

Where you really save: AI vs traditional production

When comparing AI and traditional production, the savings don't come from creativity: idea, script and direction are still the work of people. They come from everything surrounding the shoot:

• Set logistics: locations, permits, travel, set building.
• Casting and talent: models, actors, extras and their image rights.
• Reshoots: changing a scene, a season or a setting doesn't mean going back on set.
• Variants and formats: adapting a commercial into 9:16, 1:1, short and localised versions costs a fraction of the initial work.

According to BCG, video is the item that absorbs most content budgets, so that's where AI's economic impact is greatest.

The three levers of return on investment

1. Cost per content

This is the most immediate lever: how much each photo, video or variant ready for publication costs. With AI it drops above all when volumes are high, because the setup work is done once and then reused.

2. Speed

A campaign that launches in weeks instead of months catches the season, the trend or the launch earlier. The value of speed is often underestimated: late content is worth less, even if it's cheap.

3. Volume and testing

More variants mean more A/B tests, and more tests mean finding the converting creative sooner. This is the lever that weighs most on video marketing ROI, and it's the principle behind an AI Content Engine for e-commerce.

How to measure the ROI of AI content

The basic formula remains the classic one: (revenue attributed to content – cost of content) / cost of content. The key is choosing what to measure. These are the KPIs we recommend to our clients:

• Cost per published piece of content, compared with previous production.
• Time from brief to publication, in days.
• Creative performance: CTR, cost per acquisition (CPA), campaign ROAS.
• Conversion rate of product pages with and without video.
• Number of variants tested per campaign and the share of winners.
• Asset reuse: how many times the same brand system generates new content.

A practical tip: always measure a "before" and an "after" period on the same channels. Without a baseline, any ROI figure is just an opinion.

When traditional production still makes sense

AI isn't the right answer for everything. Traditional production, or a hybrid approach, is still preferable when:

• the value of the content lies in a real, recognisable person, such as a spokesperson or a founder;
• the product needs to be shown working for real, for example a machine or a technical device;
• only one piece of content is needed, with no variants or ongoing distribution: in this case the scale advantage shrinks.

In most projects the best solution is hybrid: real footage where authenticity matters, AI where settings, scale and speed are needed.

Hidden costs to include in the quote

A well-calculated ROI also includes items that are often forgotten:

• Revisions: how many are included and what extra ones cost.
• Usage rights: channels, territories and duration of use.
• Transparency: the obligations for AI-generated content under the European AI Act (Article 50).
• Quality control: human checks on product fidelity and brand consistency.

A quote that doesn't specify these items makes it impossible to compare different offers.