1 LiveRamp opened LAB — outside agents now plug into its network
On June 17, LiveRamp launched the LiveRamp Agent Builders (LAB) program, a formal initiative letting third-party AI companies bring purpose-built agents directly onto its data collaboration platform. Use cases include audience building, media analytics, cross-media measurement, first-party data prep, and activation. Founding partners: SemantIQ, Newton Research, Akkio, and Datalinx. Partners get LiveRamp's APIs and MCP servers, with access to the entire LiveRamp customer base during the pilot — not via individual commercial deals.
The interesting move is procurement, not product. Instead of forcing brands to negotiate with each AI vendor one by one, LAB lets a marketer try a portfolio of agents on the same data layer without separate contracts. That structurally changes who wins the agentic land grab: the platform that owns the connections gets to broker access to a market of agents, not pick a single winner. Watch which agent categories LiveRamp expands LAB into next — that is the roadmap of where it expects identity to remain valuable as Publicis closes its $2.2B acquisition.
2 Mediaocean lit up NIVO — 12 agents across $200B in annualized spend
Mediaocean opened the pre-Cannes run on June 11 with NIVO AI: twelve specialized agents spanning creative, delivery, measurement, and optimization, operating across platforms that already process more than $200 billion in annualized ad spend. NIVO is positioned as an autonomous layer that sits across Mediaocean's planning, buying, and reconciliation infrastructure rather than as a chat layer bolted on.
The volume is the headline. When a platform handling $200B in spend says its agents are now executing, "agentic" stops being a marketing word and starts being a workflow. The risk for brands and agencies is governance — an autonomous agent that misallocates 1% of a $50M plan is a half-million-dollar mistake nobody noticed in time. Before you let any vendor's agents touch production budgets, lock down logs, kill switches, and audit trails. Speed without observability is just expensive.
3 DoubleVerify shipped DV Neura — verification got its own AI engine
DoubleVerify launched DV Neura, an AI engine that runs across the DV Media AdVantage Platform and ties verification, content classification, optimization, and agentic execution together. Neura is positioned as the layer that lets DV measure and adapt in flight rather than report after the fact — with agents that can throttle bids, swap placements, or re-classify content as signals change.
The reason this matters: as buying agents move from advisory to autonomous, the verification layer needs to move with them — or brands lose the only seatbelt they had. DV is saying it can stay current with a buying stack that no longer waits for human approval. If you are running media at scale, the question to ask is not whether your verification vendor is "AI-powered" but whether its agents can talk to your buying agents in real time. Anything slower is now a liability.
4 Stagwell launched The Media Machine — an operating layer, not a tool
Stagwell launched The Media Machine, a media-specific operating layer that extends The Machine, the agentic platform it formally launched in January. The new layer is sold to agencies and in-house teams as media planning, buying, and optimization built around autonomous agents, with the same workflow scaffolding Stagwell uses internally across its own holding-company shops.
Holding companies have spent two years preaching agentic transformation; this is the part where one of them sells the playbook. The pitch lands particularly hard for mid-market agencies who cannot fund their own agent platforms but cannot afford to look behind, either. The buy decision is no longer "do I license an AI tool" — it is "do I run my media practice on someone else's operating layer, or build my own?" For most shops, building their own stopped being economic about six months ago.
5 Magnite Orchestration made coordination a neutral layer
Magnite launched Magnite Orchestration, a neutral coordination layer connecting buyer AI agents directly to premium omnichannel inventory without forcing buyers onto Magnite's native tools. First-named partners include dentsu and DIRECTV Advertising. The pitch: buyers bring whichever agent they like; Magnite handles the inventory side and the matchmaking, neutrally.
This is the SSP saying the quiet part out loud: in an agentic world, the seller's best move is to be the most reachable inventory layer, not the slickest UI. If buyer agents can plug in via open protocols, the platform that wins is the one that does not pick a fight with them. Worth watching whether the rest of the SSP world follows Magnite's lead or doubles down on proprietary stacks. The brands and agencies running the agents already have a preference.
6 Databricks CustomerLake landed — the agentic CDP is inside the lakehouse
Databricks formally unveiled CustomerLake at Data + AI Summit on June 16: an agentic CDP built natively into the lakehouse with no separate system or data duplication. Profile Agents and Campaign Agents drive what Databricks calls "infinity campaigns" — continuous loops of analyze, recommend, and activate across channels. Launch customers: HP, Circle K, AB InBev, and Getnet by Santander. Gartner is now forecasting 80% of net-new enterprise CDP deployments by 2030 will be embedded in or composable with a data platform.
This is infrastructure marching on the application layer, and it matters because every CDP RFP this fall just got a new line item: "lives where my data already does." Standalone CDPs aren't dead, but the burden of proof shifted — if your CDP needs you to copy data into it, you owe procurement an explanation. The pragmatic move for marketing leaders is to test whether your warehouse vendor's native activation already meets 80% of your use cases. If it does, the CDP shortlist gets a lot shorter.
7 OpenAI defined ChatGPT ad tooling — Audience Tools and Creative Tools, by policy
OpenAI published new Ad Tools Terms defining two optional advertiser features inside ChatGPT: Audience Tools (first-party data upload for matching) and Creative Tools (AI ad generation). Neither was confirmed live as of mid-week, but the legal scaffolding now exists — the same two pillars Meta and Google use to compete for performance budgets.
Defining audience matching and AI creative in lawyer-written policy is how a platform tells the market it intends to compete head-on. Brands and agencies should treat this as a "yes, this is happening" signal and start the work that takes the longest: getting first-party data clean enough to upload, and getting creative pipelines built to generate at the volume an AI native channel demands. By the time both tools turn on, the brands that prepared will already be testing and the brands that waited will be reading case studies.
8 Threads got serious about brand safety — IAS, DV, Scope3, Zefr all in
Meta expanded third-party ad-verification partnerships to Threads, giving advertisers access to content block lists and brand-safety controls from Integral Ad Science, DoubleVerify, Scope3, and Zefr. The capabilities include safety and suitability scores, content examples with associated risk levels, and impression-level data — the same primitives advertisers expect on Facebook and Instagram. Threads also added catalog ads and a drag-and-drop carousel reordering tool.
Threads has been slowly becoming a real ad surface, and "brand-safety partner coverage" is the maturity signal media buyers actually care about. Once IAS and DV are reporting on a placement, internal procurement objections start to fall away. If you have been waiting to add Threads to the plan because you could not get your suitability framework approved, the blocker just got removed. The other half of the story: a recent Adalytics report pushed transparency questions back into the brand-safety conversation. The vendors are everywhere; the audits are not. Ask harder questions of both.