For most of the past decade, running video advertising on premium channels meant three separate costs: an agency to plan it, a production team to shoot it, and a media budget large enough to clear a platform's minimum spend. A category of AI-native advertising platforms is now built on the argument that all three can be compressed into software. Smavida, which describes itself as a system that analyses a business, creates video ads and places them where its customers already spend time, is a clear example of the pitch.

Editorial note

THRYV has not tested Smavida, has no commercial relationship with the company and publishes no rating for it. Everything described below is drawn from the company's own public website, cited at the end, and is used to explain a category — not to recommend a vendor.

What the platform says it does

Smavida's public positioning has three parts. The first is what it calls autonomous creative intelligence: generating large numbers of on-brand video variations automatically, differentiated by audience, geography, intent and platform. The second is a single activation layer covering connected TV, streaming, social, retail media and digital out-of-home. The third is commercial: access to premium inventory without enterprise minimum spends or long agency contracts, with what it describes as itemised CPMs by channel and strategy, production and reporting included rather than billed as add-ons.

Its published workflow runs in four stages — planning and strategy, ad creation, activation across channels, and campaign intelligence — with each campaign's results feeding the next. That loop is the standard structure of programmatic advertising; the AI claim is about how much of it is automated and how much creative comes out the other end.

Why the category exists

Two structural shifts made this possible. Television advertising moved onto internet-delivered, addressable inventory, so a small advertiser can in principle buy the same streaming placement as a national brand in far smaller increments. At the same time, generative video tooling cut the marginal cost of producing a fifteenth variation of an advert close to zero. Where the old constraint was production capacity, the new constraint is deciding which variation to run and how to tell whether it worked.

Generating a thousand ad variations is a solved problem. Knowing which one earned the sale is not.
THRYV Tech Desk

The questions that decide the outcome

1. Who owns the creative, and how good is it?

Ask to see finished output from a business in your own sector, not a sizzle reel. Establish in writing whether you own the generated assets, whether they can be reused outside the platform, and who is liable if a generated advert misrepresents your product or breaches an advertising code.

2. What exactly are you paying for?

An all-inclusive price is only transparent if the media cost is itemised separately from the platform's own margin. Ask for CPMs by channel, the platform fee expressed as a percentage of media spend, and confirmation of whether the company is buying as an agent on your behalf or reselling inventory it has bought itself. The two arrangements create very different incentives.

3. How is performance measured?

Connected TV and out-of-home do not produce clicks. Reported outcomes therefore depend on attribution modelling, and the vendor usually supplies the model. Ask which measurement is independent, whether you receive log-level or placement-level delivery data, and whether the platform will run a holdout test — a matched audience that sees no advertising — as the honest baseline.

4. Where does your data go?

Business inputs, customer lists and campaign results are all processed to make the system work. If you operate in the EU or UK, establish the lawful basis, the processor agreements and whether your data trains models used for other advertisers.

A short due-diligence list

  • Request three finished creatives and one full delivery report from a comparable advertiser, with permission to contact them.
  • Get the media cost and the platform fee on separate lines of the same document.
  • Agree a minimum test budget and a fixed test window before committing to any longer term.
  • Define one primary business outcome — bookings, qualified leads, revenue — and one holdout to measure it against.
  • Confirm asset ownership, contract exit terms and data-processing arrangements in writing.

None of this is specific to Smavida. It is the standard the whole AI-advertising category should be held to, and it is the standard a small business needs precisely because the pitch removes the agency layer that used to ask these questions on its behalf.