Is AI helping agencies break out?
The conventional wisdom, which I’ve certainly espoused, is that ad agencies are essentially big bundles of people doing knowledge work, and that AI is inevitably, and quite obviously, going to cut deeply into that offering.
That wisdom is still likely true, but how it is happening feels a little surprising. One of my hot takes from Cannes is that agencies might be using AI to actively dsitrupt themselves rather than wait for tech to do it for them. With the important caveat that IT’S SOOOOO EARLY, and we don’t really know what’s going to happen, one of the distinct themes from Cannes was that agencies are actually shipping code, not just talking shit.
AI is a more fundamental technology than previous waves
With previous waves of technology rolling through the advertising world the role of the agency customer has been one of adoption, coordination and expertise. Whether we’re talking about ad serving, DMPs, programmatic, analytics, or the rest of the tech stack that’s made the enormous growth of digital advertising possible, the agency was, at best, a smart and sophisticated customer.
To the extent that agencies were driving the technology forward, it was usually as a power user of a vendor’s offering, perhaps as a white labeled instance. Agency Trading Desks, for example, were often the biggest and most powerful customers of DSPs, but rarely built their own. Agencies might partner with vendors to innovate on behalf of one or more customers, then give the output a cool name like “Platform XYZ” and use it to acquire customers.
My contention is that the AI wave is a more fundamental technology than any application layer ad tech solution we’ve seen previously, and is prompting (pun intended) agencies to build solutions themselves, rather than wait for solutions to emerge from entrepreneurs and vendors. AI is more akin to general computing than it is to any advertising specific technology.
Some interesting data points
On Tuesday of last week I moderated a panel with Steve Carbone from Orion Worldwide (an IPG Company) and Marion Hargett, the Chief Revenue Officer of MiQ. Let’s highlight what both of these companies are up to:
IPG’s has various investments in AI, including a partnership with Adobe and their Interact product. This week they announced a partnership with Snowflake which I find pretty interesting, because it allows customers to use the data intelligence they’ve built without moving data into the agency’s cloud.
MiQ (though definitely not an agency), has been making waves with its new Sigma product that promises to create AI-driven personas then use AI to programmatically reach them.
On Thursday I interviewed old friend Oleg Korenfeld (pod link) from CMI Media, who walked me through the development of an AI media planner that is in production for traders at this health-focused division of WPP.
In previous newsletters and pods we’ve spoken about WPP’s CTO (and friend of the pod) Stephan Pretorius, who is featured in this Bloomberg video giving demos of the various products the holding co built using AI.
I also got a chance to interview Crissi Cupak, the Head of Product for PMG, about their product, called Alli. Alli is pitched as a “Marketing Operating System”, which I can’t directly vouch for, but last week they announced a marketplace product which brings data and capabilities from 25 vendors into the suite. This is a long way away from the more traditional model of agency-preferred vendors, or recommended partners.
There are a lot more examples out there of agencies creating real products that are in the hands of their clients and internal teams. It is a bit of a stark contrast to the number of start-ups and established ad tech vendors that are talking “agentic” but largely haven’t shipped anything yet, two years into the LLM revolution.
Core vs non-core technologies
Ultimately companies are going to invest in technology where they think they can gain an advantage by doing so. That seems obvious. In the past waves of tech, agencies have put money into the area of data (Xaxis, etc), while eschewing investments in activation (DSPs). While it is TOO EARLY to tell where things are going to shake out, some early signs of the AI investment landscape:
Creatives: Agencies are partnering with Adobe and other AI vendors, then making the offering bespoke for their specific needs.
Media planning: Building directly as it is core to the agency value proposition and the data is too sensitive to share.
Orchestration: Build directly, same reason as above.
Measurement: TBD, I don’t see enough evidence to come to a conclusion.
Activation / targeting: Leave it to the specialists like MiQ, TTD, etc.
Agree or disagree? Let me know your thoughts.
Reading list
Reading list will be back next week.