In this op-ed, Luc Benyon takes a look at agentic age, and the barter deal we cut when we trade in vibes...

It seems the era of agentic advertising has officially begun. With the launch of Ad Context Protocol, AI-powered ad trading is no longer a novelty – it’s a paradigm shift. It’s strange, because I had thought we were in the 'outcomes era'. In fact, 'outcomes' are a currency, and AI is a technology – so the two eras can exist simultaneously. But what does this look like?

First came 'vibe coding': developer shorthand for the way applications could be written intuitively, conversationally, using AI interfaces. The idea caught the collective imagination, and now vibes are all the rage: in the world of ad tech, the concept of 'Vibe Targeting' has emerged.

Vibe targeting

Vibe targeting, as conceived by Brian O’Kelley and explored by Shirley Marschall, is all about finding audiences to advertise to, but with a general feel of who they should be, rather than specifics.

Vibes are the way we interface with AI, using prompts: a method of communication that retains our human desire to converse, speculate, and brain-dump, while instructing a machine to undertake an extremely complex task. It’s trial and error, it’s experimental, it’s… vibes.

With vibe targeting, we skip the rigid taxonomy of the DSP interface, and describe our ideal audience: their demographics, behaviours, and interests (BOK calls this our 'Brand Story' – and naturally, so does the Scope 3 interface). Then we let the DSP do the rest. Like much of the impact of AI, this is essentially replacing an antiquated interface of check boxes with a more conversational or 'vibey' one.

Our vibe targeting input could theoretically be cut and pasted directly from the client brief to their agency.

There have been plenty of innovations around the creation of audiences in recent years. Often inspired by the need to find cookieless solutions, these audience creation tools aggregate data from classic and non-traditional signals agnostically. These could be web consumption behaviour, social media activity, or TV viewership, alongside contextual signals. 

The open-endedness of the vibe targeting approach lends itself well to these data agnostic demand side entry points. Put simply: you describe the persona or audience you want to reach, and then let the platform do the rest. Adding AI-powered conversational interfaces to these products (like StackAdapt’s Ivy or Nexxen’s NexAI) opens the door for audience creation to be more conversational and conceptual.

For Shirley Marschall (writing in ExchangeWire), though this leads to an accountability problem, "The vibe feels good. The campaign 'performs'. But no one can explain how or why. We’ve moved from precision to opacity. From creepy to clueless. From strategy… to vibes." 

Marschall is right: with each step away from selecting inventory, demographics or segments, we’re removing our oversight of our campaigns – and thus our control. That is the barter deal we cut when we trade in vibes. 

The outcomes era

Let’s shift focus from inputs for a second, and consider the outputs of our advertising campaigns. 

In the outcomes era thesis (of which Joe Zappa is the most prominent proponent), we buy media using the specific outcome as the currency; pounds or dollars are exchanged for sales uplift, app installs, or basket-fills. 

The outcomes era concept has been driven by digital technology, enabling increasingly downstream attribution. From a buyer’s perspective, this is great, as they move beyond vanity and proxy metrics and increasingly towards real attribution.

Once the province of social and native advertising, then display, this idea is currently even creeping increasingly into video and CTV buys. And it’s here where it has come up against some resistance. Video, and in particular, CTV, is traditionally a brand building medium, not performance; and while brand building and sales both drive growth, they require different approaches.

Vibing outcomes 

This brings us back to our vibe targeting concept. Can you 'vibe' outcomes?

There’s no reason why not.

The initial benefit of vibe coding or vibe targeting is that you don’t need to be proficient in Java Script, or data taxonomies (respectively). So the inference is that it’s board-strokes, vague even. 

But even the vaguest of campaign managers knows what they want out of a campaign. Any decent brief will begin with a business goal. Why shouldn’t we be able to vibe our business goal, and let the technology do the work to deliver us an outcome-based campaign? 

Currently, we extract who we believe to be the audiences that we want to convert, and then map those against segments built from a combination of behaviours, interests (derived from behaviours) and demographics (likely derived from behaviours). At each step of creating these 'audiences', we’re making assumptions and betting on the lowest common denominators. 

This is something AI can do better than us. In fact, it’s a no-brainer to offload the task of creative and optimise audience buys to AI. Data at scale is something AI is made to do.

And as for our brand-building campaigns, well – this is where attribution has failed to yet provide a satisfactory solution. Perhaps vibes can become an output for brand campaigns: A 20% uplift in 'vibe recognition' or 'vibe uplift', anyone? There’s work to be done here, clearly.

The combination of AI Interface + Signal-Based Audience Creation + Programmatic Delivery Pipes can translate vibes into outcomes. In fact, it is doing so in demand-side platforms as we speak. It requires new skills in what we ask of our platforms and what we put into them. But ultimately, it should make the job of audience creation easier – and connect audiences to tangible business benefits more convincingly.