Marketing Weekly AI News
September 21 - September 29, 2026Weekly signal
From Sep 21–29, 2026 the industry shipped the pieces that decide whether agentic marketing becomes a new channel you buy into or a technical experiment you audit. Product and platform vendors focused on the interfaces between humans, brand-owned data, and autonomous agents: ad platforms turned clicks into brand-run conversations; standards bodies published an agent-to-agent RFP/proposal spec; and vendors and vertical incumbents began running revenue workflows through agentic applications. Together these moves shift the work marketers must do—from copy and targeting to catalog hygiene, machine-readable briefs, and agent governance.
What changed
OpenAI: Sponsored Agents, Shopify & HubSpot integrations
OpenAI formalized a new advertising experience that lets a user who clicks an ad open an explicitly labelled conversation with an advertiser-run agent ("Sponsored Agents") and expanded its Ads Manager with AI-assisted creative tools. The company also shipped a ChatGPT Ads app that reads Shopify product catalogs so U.S. merchants can create and manage ChatGPT Ads from Shopify; OpenAI said international markets follow from Sep 23. Practically, that means paid placements can now hand a user directly into a controlled conversational surface where the brand's agent guides discovery, answers questions, and routes to checkout. This changes both creative (conversations, decision trees) and measurement (agents create a new funnel stage).
IAB Tech Lab: AAMP 3.0 / OpenProposal (RFP→Buy)
IAB Tech Lab published AAMP 3.0 which adds OpenProposal, a specification for how sellers represent their inventory and how buyers present briefs so that buyer and seller agents can discover, evaluate and compare proposals automatically. AAMP 3.0 is explicitly aimed at standardizing the historically manual stage between an RFP and a buy commitment—exactly the step that will be automated as agents take on planning and negotiation tasks. The spec is in public comment, which is the window for vendors and enterprise buyers to shape the semantics and enforcement model.
Vendor & vertical deployments: Kana + TIME; Vertafore
Kana announced that TIME has deployed its agentic marketing applications for ad sales and campaign delivery; this is a visible publisher use case where proposal turnaround and campaign ops are being compressed by agents. In the insurance vertical, Vertafore unveiled an "Agentic Agency" vision with new Velocity AI agents embedded in distribution workflows. Together these show the same pattern across industries: agents moving from pilot to actual commercial workflows and revenue operations. That makes the operational questions—data fidelity, auditability, and seller-side schema—urgent.
Supporting protocols: Agentic Audiences / embedding exchanges
Separately, open-source/spec work (Agentic Audiences) is progressing to define how agents exchange learned embeddings that encode identity, context, and reinforcement signals. That work matters because it defines what 'audiences' mean to an agentic market and how measurement and privacy controls are carried forward across agent interactions.
Why this matters for marketing teams (implications)
-
New funnel stage. Agent conversations are a measurable post-click event. That requires new KPIs (conversation-qualified lead, time-to-hand-off, agent drop-off reasons) and updated attribution models that separate impression→click from Sponsored-Agent engagement→conversion. Expect early campaigns to show high engagement but mixed ROAS until tracking, billing and measurement converge.
-
Catalog & schema are now the moat. Agents treat structured catalog data and proposal descriptors as the primary signal for discoverability and matching. If your product feed, pricing, inventory, or policy metadata is incomplete, agents will not select you—even if your marketing or SEO is excellent. Prioritize canonical, machine-readable product data and a content-to-schema mapping project.
-
Standards matter operationally. AAMP/OpenProposal and Agentic Audiences aim to make agentic buying repeatable across publishers and marketplaces. Teams that ignore these specs will face custom integrations and brittle automation. Early engagement with IAB Tech Lab public comment and vendor roadmaps will reduce rework and vendor lock-in.
-
Governance & disclosure risk. Sponsored Agents are explicitly labelled, but brands must control what the agent can say, what data it uses, and how it routes users to conversion. Legal, privacy and brand safety teams must own an operational checklist (prompts, data sources, retention, handoff rules, audit logs). Regulators are watching agent-driven commerce closely; keep transparency controls and human escalation points.
Practical next steps (what to do this week)
-
Run a 30–60 day pilot: pick a narrow catalog (top SKUs), install the ChatGPT Ads Shopify connector (if you’re a Shopify merchant / partner) or connect via the HubSpot integration if you manage B2B leads. Measure agent-engaged conversions as separate funnel events. Budget small, but instrument thoroughly.
-
Clean your feed: audit product metadata (title, description, size/fit, inventory, price, return policy, canonical image tags). Create a prioritized remediation backlog—agents prefer accuracy, not cleverness.
-
Prepare machine-readable briefs: start converting one typical RFP/brief into a JSON/agent-friendly format, and test how a selling agent would parse it. Engage vendor partners on AAMP/OpenProposal compatibility. If you buy media programmatically, ask your DSP and publisher partners about OpenProposal and Agentic Audiences support.
-
Establish an agent governance playbook: define allowed responses, escalation triggers, data retention, and logging. Capture prompts and back-end evidence used by the agent for compliance/audit. Tie this to your existing ad policy and privacy reviews.
-
Engage standards and vendor forums: AAMP 3.0 is in public comment—send product/engineering feedback if your stack will transact via agents. Track vendor roadmaps for OpenProposal/AdCP and Agentic Audiences conformance.
Closing note
This week’s activity shows the channel is moving fast from demos to commerce-ready pieces: if you run advertising or commerce, you should shift a small amount of budget and ops attention to agentic experiments—focused on catalog quality, measurement instrumentation, and governance—before adoption pockets harden around particular standards and partner implementations.
Stop reading agent demos. Give one a job you repeat every week.
Describe the work, test the first result, and keep the agent available without running your own server.
Plans from $130/month. Cancel anytime.
Hosted agent
OpenClaw or Hermes