Agency AI Agents Threaten to Accelerate Client In-Housing of Media Buying
The Development
Dentsu has entered a partnership with Newton Research, an AI agent development company, to embed transactional media-buying agents and faster analytics capabilities into its Dentsu.connect platform. Caitlin Gelles, EVP of data technology and measurement at Dentsu, confirmed active early-stage pilots with U.S. clients targeting transactional buying automation and accelerated campaign setup. The announcement lands alongside WPP's parallel disclosure that it is testing a media buying and planning agent for premium video inventory. Both moves are being positioned as agency-side capability plays at Cannes Lions 2026, where AI agent demonstrations have dominated holding company messaging.
Our Take
The strategic contradiction at the centre of this moment is being underplayed. The same AI agent infrastructure that Dentsu and WPP are building to defend margin and demonstrate value is, by design, legible and reproducible. Any client with a capable in-house data team and a Newton Research or equivalent vendor relationship can replicate the transactional buying layer without agency intermediation. Agencies are training their own replacements and calling it a partnership model. The genuine differentiator — if one exists — is proprietary data, preferential inventory access, and cross-client pattern recognition. Neither Dentsu nor WPP has made a credible case that their agent deployments are built on anything clients cannot eventually procure directly.
What Changed
Agencies can now deploy autonomous agents to execute transactional media buys — not just plan or recommend them. This shifts the agency's role from execution to governance, compressing the labour-intensive middle layer of the campaign lifecycle that has historically justified retainer and fee structures.
Marketing Impact
Media buying and marketing operations teams inside large advertisers are the primary audience. The technology now exists to bring transactional programmatic execution in-house at lower operational cost — the barrier is organisational will and data infrastructure readiness, not technology access.
Competitive Implication
Holding companies with proprietary data assets and platform-level inventory relationships retain short-term advantage. Independent agencies and mid-tier holding company units without differentiated data layers become structurally vulnerable as clients benchmark agent-driven costs against agency fees.
Strategic Outlook
Expect two to three major advertiser in-housing announcements citing AI agent capability by Q1 2027. Agencies will respond by accelerating proprietary data partnerships and repositioning toward strategic governance — but the fee compression on execution will be permanent once client pilots demonstrate cost parity.
The Exploit
Action Item
Heads of media at brands currently spending above $50M in programmatic should issue an RFI to Newton Research or a comparable AI agent vendor before Q3 2026 budget cycles close, establishing a baseline cost comparison against current agency execution fees.