1 AdRoll wired Claude and ChatGPT directly into its ad data
AdRoll launched an MCP server connecting its advertising data repository directly with external AI applications including ChatGPT and Claude. Users write a prompt in their preferred LLM, and AdRoll responds with performance reports, audience insights, and campaign management actions — without opening the AdRoll dashboard.
This is what "MCP as the new ad API" actually looks like in practice. AdRoll didn't build a new agent; it exposed its existing data through a protocol the LLMs already speak. The strategic move for buyers: stop scoring vendors on "do they have AI?" and start scoring them on "do they have MCP?" If your DSP, attribution tool, or ESP can't be queried from the chat you already use, your team is doing dashboard hopping instead of media planning.
2 Netcore took SQL out of audience building
Netcore launched Audience Agent, a segmentation system that interprets plain-language descriptions to build customer lists. "Customers who bought baby gear in the last 90 days but haven't logged in in 30" returns a list, with no SQL, no data team ticket, no waiting. The agent reasons over campaign structures, audience definitions, and product catalogs in real time.
Audience building has been a bottleneck since CDPs got popular — marketers know the segment they want; the data team is six tickets deep. The natural-language layer collapses that delay to seconds. The asterisk: the agent is only as good as the schema underneath. If your CDP's data model is messy, a chat-based segmentation tool will be confidently wrong. Clean the foundations, then turn on the language layer. The order matters.
3 Bloomreach made retention autonomous
Bloomreach released an AI agent that lowers e-commerce churn by tracking shopper behavior, identifying disengagement signals, and autonomously dispatching targeted SMS or email discounts to re-engage at-risk customers. The agent identifies, selects an intervention, and delivers it — no manual campaign setup.
This is the structural difference between marketing automation and agentic marketing. Automation runs rules you wrote; agents make the call. The win is speed; the risk is that the agent's idea of "at-risk" might be confidently miscalibrated against your actual margin model. Pilot in a contained segment with a clear hold-out group, and measure incrementality, not save rate. "Saves" alone reward over-discounting; incrementality tells you whether the agent earned the spend.
4 DoorDash relabeled its ad business a "global commerce media platform"
DoorDash announced the transformation of its advertising business into a global commerce media platform, surfacing targeted ads to shoppers across the app and external surfaces. The rename matters: "commerce media" signals DoorDash wants to be on the same procurement RFPs as Amazon Ads, Walmart Connect, and Instacart Ads — not just an in-app sponsored placement.
For brands selling food, household, and beverage, DoorDash now has the dataset (basket, location, frequency, drop-time) to compete on the same closed-loop pitch as the big retail media networks. The platform also has a real differentiator: timing. DoorDash knows when you ordered, not just what you bought. Pair that with the ads, and you get the closest thing to a real-time consumption signal for CPG. Test it before everyone else does.
5 InMobi and Scope3 shipped a sell-side agent
InMobi Advertising, in partnership with Scope3, launched a programmatic ad system that opens mobile and television ad spaces directly to autonomous AI buyers. The platform interfaces with machine-learning purchasing software that reads pricing and makes decisions without human intervention — a sell-side agent designed to negotiate with buy-side agents.
This is the logical endpoint of programmatic: agent on the buy side, agent on the sell side, humans on exceptions. For marketers, the new question is governance — what's the equivalent of an inclusion list when your agent is negotiating with the seller's agent? The IAB and Ad Context Protocol working groups are racing to write that vocabulary. In the meantime, push your DSPs and SSPs for transparency on what their agents are agreeing to on your behalf.
6 OpenAI rolled out conversion-optimised ChatGPT campaigns
OpenAI expanded its advertising platform with conversion-optimised campaign types, using pixel and server-side measurement tools to track actions like purchases and sign-ups. The Conversions API mirrors what's been table stakes on Meta and Google for a decade — and finally lets advertisers buy into ChatGPT with the same KPIs they buy anywhere else.
Until this week, ChatGPT ads were closer to a brand bet than a performance one — there wasn't a clean way to optimise against revenue. Conversion campaigns close that gap. Expect ChatGPT spend to move out of the experimental innovation budget and into the performance team's testing queue. The shape of the next twelve months: ChatGPT becomes a real bid line item alongside Google Performance Max and Meta Advantage+.
7 ANA Masters of B2B handed out the 2026 B2 Awards
The ANA Masters of B2B Marketing Conference wrapped its main programming day at the Chicago Marriott on the Magnificent Mile, with the 2026 B2 Awards announced this evening. The conference brought together senior marketing leaders, ABM strategists, demand-gen heads and analytics chiefs — the largest single B2B gathering in the US calendar.
The B2 Awards aren't just industry pats on the back — they're the only awards bench-marked entirely on B2B outcomes (pipeline, ABM efficiency, lifetime value). If you sell into B2B and you're staffing the next quarter, the winners' list is the most credible recruiting source you have. It's also the cleanest signal on which agencies are actually moving B2B numbers in 2026, separating real practice from rebadged consumer playbooks.
8 Marketers got natural-language tools they actually wanted
Reading AdRoll, Netcore and Bloomreach together, the unifying theme is unmistakable: SQL is being pulled out of marketing. Three different vendors shipped three different surfaces in the same week, all rooted in the same insight: the language a marketer thinks in is English, not query syntax.
This is the maturation marketers have asked for since the first CDP demo — "why can't I just ask?" The catch is the discipline most marketing teams haven't built: the schema, taxonomy, and naming conventions that make natural-language queries reliable. If your team wants to use these tools well, the work is still in the data layer, not in the UI. Spend the next two weeks on a clean event taxonomy; you'll get the natural-language payoff for years.
9 Apple's WWDC preview reset on-device AI for marketers
Earlier in the week, Apple's WWDC 2026 introduced a redesigned Siri powered by Apple Intelligence, with stronger conversational capability, deeper app integration, and a combination of on-device and cloud-based processing. The marketing-stack implication: on-device AI now has a credible consumer surface that competes with Google's Gemini and OpenAI's ChatGPT as the user's default assistant.
For brands, the practical impact comes in iOS app marketing and Siri-surfaced shortcuts. If the assistant can act inside your app, you have a new acquisition and retention loop — if it can't, you're a less convenient choice. Audit your iOS app's intent surface this quarter; the assistant marketplace is being built in front of you, and being absent from it is a default loss.
10 The week's theme: AI moved from feature to infrastructure
Pull the launches together — AdRoll MCP, Netcore Audience Agent, Bloomreach churn agent, DoorDash commerce media, InMobi/Scope3 sell-side agent — and the same line shows up under each: AI isn't the product anymore; it's the protocol the product talks to. "AI feature" is becoming as redundant as "web-based feature" was in 2008.
The implication for budget allocation is direct: stop earmarking a separate "AI experimentation" line. Embed the AI evaluation into every renewal — CDP, DSP, ESP, attribution, BI, support — with a single criterion: does this tool expose its data through an LLM-friendly protocol, or am I still paying for someone else's dashboard? The answer tells you which vendors are wiring themselves into the next decade, and which ones are quietly opting out.