Just in: What Retail Media's US$200 Billion Ceiling Means for Brands Selling Through AI Agents

Is retail media's growth story over?

Not over — but the easy part is. WARC's Future of Commerce Media 2026 report puts global retail media spend past US$200 billion this year, heading to US$223.4 billion in 2027 and around 15.2 per cent of total global ad investment. Still the fastest-growing corner of advertising. But WARC is blunt that the expansion rate is cooling toward single digits after several years of outsized growth, and Amazon, Walmart and Instacart alone hold 78 per cent of the category between them.

So what for brands: you've spent three or four years treating retail media as the reliable line item that always grew. That assumption is now shakier, at the exact moment AI agents are starting to intermediate how products get found in the first place. Two structural shifts landing on the same budget line, not in sequence. The brands that get ahead here are the ones asking their top retail media partners a pointed question this quarter: how does our inventory actually surface inside agent-led discovery, not just human scrolling.

Why is agent security suddenly the bottleneck, not the tech?

Google Cloud's latest State of AI Infrastructure commentary names agent security as the top gating issue standing between agentic pilots and real production scale, and recommends secure AI frameworks, platform-level governance, task-level provenance and human-in-the-loop checks before agents get real budget and real customer data.

This is worth sitting with, because most of the coverage this year has been about protocols — ACP, UCP, Agent Pay, checkout flows. Useful, but it skips the harder question: does your board actually trust an autonomous agent to spend money or touch a customer record unsupervised? Most plausibly, that trust gap — not the underlying capability — is what determines how fast agentic commerce scales inside any given enterprise over the next twelve months. If your organisation doesn't have a named owner for agent spend limits and provenance yet, that's the gap worth closing before the next budget cycle forces the issue.

What does Australia's agentic banking rollout tell brands elsewhere?

Mi3's reporting on Westpac, Tower, CBA, NAB, ASX and Barrenjoey shows agentic AI has left the pilot phase entirely in Australian banking. Westpac's agents are processing payslips and lending checks and returning more than 150,000 employee hours a year. CBA took part in Australia's first authenticated agentic payment transactions on Mastercard's rails. NAB is deliberately building safeguards before scaling further, rather than after. Alongside that, Salesforce's newest ANZ commerce data found 91 per cent of Australian commerce leaders now agree large language models will be essential to product discovery within a year, with AI-assisted discovery up 38 per cent since last August while traditional search referrals fell 15 per cent.

This is the clearest real-world proof point going: agentic AI running regulated, high-stakes processes at genuine scale, not chatbot theatre. For brands outside financial services, the discovery numbers are the sharper signal — the channel mix that built your funnel is already losing share, and the shift is measurable, not theoretical.

Are the ad platforms building agents to buy media, not just sell it?

Over the past couple of months, OpenAI has rolled out conversion-optimised ChatGPT ad campaigns with automatic budget pacing, and Meta has opened Ads Manager to AI agents through its own MCP server, working toward full ad automation from a single URL or product image — creative, audience and placement generated without a human at the console.

This is the loop most people miss. It isn't only that AI agents will shop on your customers' behalf. It's that AI agents may soon be buying the media that reaches those customers too. Marketing is being disintermediated from both ends of the funnel at the same time, and the sensible move is getting your media team hands-on with agent-operable tools now, while the stakes are still small.

The takeaway

Put these four together and the picture is consistent: the infrastructure layer of agentic commerce — protocols, payment rails, checkout flows — is largely built. The contest has shifted up a level, to trust, governance and who controls the budgets that fund discovery in a world where agents do a growing share of the looking and the buying. If you take one thing into your next leadership meeting, make it this: audit where your brand shows up when an agent does the searching, not just when a person does. That single exercise will tell you more about your Q4 exposure than another slide on protocol standards ever will.