1 Salesforce bought Fin for $3.6B — the agent is now the acquisition target
On June 15, Salesforce signed a definitive agreement to acquire Fin — the AI customer-service company formerly known as Intercom — for roughly $3.6 billion, folding Fin's "customer agent" into Agentforce. Fin claims its proprietary model, Apex, resolves 76% of support volume end-to-end across chat, email, WhatsApp, SMS, phone, and Slack — and brings more than 30,000 business customers with it. The deal is expected to close in Q4 FY27 with no change to Salesforce's FY27 guidance.
The signal: customer-service AI just stopped being a feature you bolt onto a CRM and became the asset a CRM pays $3.6B to own. Every vendor with an "AI assistant" on a customer-support roadmap should read this as both an opportunity and a warning. The opportunity is that resolution rate — not seat count — is now the metric that gets you bought. The warning is that the incumbents are no longer waiting for you to scale; they are buying the resolution layer directly. If your support stack is still measured by tickets touched, you are measuring the wrong thing.
2 Cordial went headless — the platform agents will actually build on
Cordial — the enterprise marketing platform behind Levi's, Tapestry, L.L.Bean, and Boot Barn — launched a headless, LLM-agnostic AI infrastructure that exposes every Cordial capability as a service any agent can call. The bundle includes an MCP server, a CLI, an API, and developer tooling, so external agents (Claude, Gemini, ChatGPT, or a brand's own) get the same access as agents running inside Cordial. CEO Jeremy Swift framed it as "headless or walled": the next era goes to the platform agents can actually build on.
This is the practical answer to the question every enterprise marketer has been asking quietly: do I have to commit to one vendor's agent stack, or can I orchestrate across the one I already use? Cordial just made the bet that openness wins, which puts pressure on every CDP, ESP, and campaign platform that still treats its agents as a moat. If you are evaluating martech in the back half of 2026, "does it expose its capabilities to outside agents?" should be question one, not five. Walled gardens are about to look very expensive.
3 Gray Media handed 117 markets to Madhive — local TV got an AI DSP
Gray Media — the largest owner of top-rated local TV stations in the US — deployed Madhive's AI-powered demand-side platform across 117 markets, with a refreshed Maverick AI engine sitting on top. The deal opens advertiser access to political audiences via L2, Tunnl, Data Trust, and TargetSmart, plus premium streaming inventory from Disney, Paramount, Roku, Tubi, and Warner Bros. Discovery, and includes in-game live sports. The timing is not subtle: it lands ahead of the 2026 midterm election cycle.
For marketers who write off local broadcast as a legacy buy, this is the moment to look again. Local TV with an AI DSP layered on top is no longer "expensive linear" — it is targeted streaming inventory with political audience data and live-sports adjacencies, sold programmatically. The midterm timing also tells you who's paying for the platform to scale: every campaign and PAC about to spend on persuasion in those 117 markets. If you are a regional brand or category that benefits from being on screen next to political and sports content, this is a buy window worth modeling, not ignoring.
4 LinkedIn split impressions in two — in-network vs. out-of-network reach
LinkedIn began a global rollout of a new post analytics view that splits impressions into in-network reach (your existing connections and followers) and out-of-network reach (everyone the algorithm decided to surface you to). The same update made saves and sends primary engagement signals over impressions, and continued the platform's penalty on posts with external URLs in the body — with some marketers reporting 50–70% lower reach when a link sits in the post itself.
The split is more useful than it looks. In-network reach tells you whether your existing audience is still reading; out-of-network reach tells you whether the algorithm is willing to put you in front of new buyers. For B2B teams, those are two different problems with two different fixes — and you have not been able to see them separately until now. Move links to the comments, optimize for saves and sends rather than likes, and start reporting in-network and out-of-network as separate KPIs. The platform just told you, in its own analytics, what it actually rewards.
5 Databricks moved into martech — CustomerLake is a CDP without the export
At Data + AI Summit on June 16, Databricks unveiled CustomerLake, an agentic CDP built natively inside the Databricks lakehouse — no separate system, no data duplication. Profile Agents and Campaign Agents power what Databricks calls "infinity campaigns": continuous, agent-driven loops that analyze, recommend, and activate across channels in real time. Launch customers include HP, Circle K, AB InBev, and Getnet by Santander. Gartner is now forecasting that by 2030, 80% of net-new enterprise CDP deployments will be embedded in or composable with a data platform.
This is the application layer marching into the infrastructure layer — or, depending on how you read it, the infrastructure layer eating the application layer for breakfast. Standalone CDPs spent a decade arguing they were the source of truth; Databricks just argued that the lakehouse already is. The practical question for any team buying a CDP this year: are you really shopping for "customer data platform," or for "an activation layer that lives next to my warehouse"? The answer changes which vendors stay on the short list and which get cut.
6 ChatGPT crossed 900M weekly users — and 20% of queries have commercial intent
OpenAI confirmed ChatGPT now has more than 900 million weekly active users, with leadership saying roughly 20% of queries already carry direct commercial intent — "where to buy," "best of," product research, travel planning. Sponsored links at the bottom of responses have now expanded to seven countries (US, UK, Canada, Australia, New Zealand, Japan, South Korea) with Brazil and Mexico queued up next. The strongest-performing verticals so far: travel, retail, health, beauty, financial services.
For anyone still treating ChatGPT as a research tool rather than a discovery surface, the numbers no longer let you. One in five queries is a buyer in motion. The job for brand teams is now twofold: (a) get cited inside the answers (organic GEO), and (b) figure out where you fit when sponsored placement opens up in your country. Treat ChatGPT visibility the way you treated Google rank ten years ago — not "if," but "how soon." Waiting until your competitor shows up next to your category is the most expensive option on the menu.
7 Meta beta'd a business AI assistant inside Ads Manager — for every advertiser, worldwide
Meta moved its in-product business AI assistant inside Ads Manager into beta for every advertiser and agency worldwide. The assistant answers natural-language questions about campaigns, surfaces optimization recommendations, and is paired with new AI Connectors that let advertisers manage Meta ads through external AI platforms and workflow tools. In parallel, Threads gained two new ad formats (static carousels and video ads) and brand-safety partnerships with Integral Ad Science, DoubleVerify, and Scope3.
The Ads Manager assistant looks small until you notice the connectors. Meta is admitting that agencies and in-house teams have already built workflows around third-party AI — Claude, ChatGPT, internal copilots — and is choosing to be reachable from inside those workflows rather than forcing everyone back into its own UI. That is the new posture for ad platforms: be a tool that agents can use, or be the tool that gets bypassed. Pick one.
8 AI skills are pulling 62% wage premiums — and jobs grew 8x faster than the rest
Lightcast's mid-year labor data showed jobs requiring specific AI skills growing nearly 8x faster than the total US job market, with the average wage premium for AI skills now at 62%. That premium has been climbing every quarter since 2024, and the demand is no longer concentrated in engineering — marketing, ops, finance, and legal are all hiring for AI-fluent roles at meaningful multiples over their non-AI equivalents.
For marketing leaders, this is a budget conversation as much as a talent one. The cost of an AI-fluent senior marketer in 2026 is materially higher than it was 18 months ago, and the gap will keep widening through the back half of the year. If you are not training the team you already have on the agents and tools you already pay for, you are about to be outbid for the talent that already knows. Internal upskilling is not a "nice to have" line in next year's plan — it is the cheaper of two unattractive options.