---
title: "Just in: What The Trade Desk Going Agentic Means for Brands Selling Through AI Agents"
description: "The Trade Desk's Ask Koa, a neutral home for AI agent protocols, and Woolworths' agentic rollout all point the same direction this week."
date: 2026-08-27
author: "MARK"
canonical: https://marksuite.ai/insights/trade-desk-ask-koa-agentic-brands
categories: ["Agentic commerce","Market data"]
publisher: MARK (https://marksuite.ai)
---

# Just in: What The Trade Desk Going Agentic Means for Brands Selling Through AI Agents

*Just in — 2026-08-27*

Four stories broke in the space of a week that, on their own, look like routine adtech and retail news. Read together, they say something sharper: the infrastructure of agentic commerce — the protocols, the interfaces, the guardrails — stopped being a slide in someone's 2027 roadmap and started shipping.


## What did The Trade Desk actually announce with Ask Koa?

The Trade Desk unveiled Ask Koa, a conversational layer sitting across its newly rebadged Kokai Zuma platform, moving from alpha to closed beta this week. Instead of a media buyer clicking through tabs to build an audience, launch a campaign or work out why a line item underperformed, they ask, in plain language, and an agent handles the workflow underneath. Early results shown to agencies point to a meaningful lift in cost-per-acquisition once the platform's forecasting engine is doing more of the decision-making, and Stagwell has already committed to piloting it (source: AdExchanger, Adweek, MediaPost).
So what for brands: this is the largest independent buying platform in the market declaring that the interface of the future is a conversation, not a dashboard. That puts pressure on every walled garden and every in-house media team to answer the same question, because the alternative is looking like the platform still stuck with spreadsheets. For brands, the practical implication is that your agency's reporting cadence — the weekly call, the quarterly deck — is quietly becoming a job description an agent can do faster. Worth asking your agency now how they plan to use it, rather than waiting for the case study.

## Why does handing A2A to a neutral foundation matter for your agentic roadmap?

A quieter but arguably more consequential development: Google has transferred governance of its Agent2Agent protocol to the Agentic AI Foundation, the same neutral body that already houses Anthropic's Model Context Protocol. MCP governs how an agent talks to its own tools; A2A governs how one agent talks to a completely different agent built on a different model by a different company. Putting both under one roof, backed by dozens of major technology partners, is what allows a shopping agent built on one company's model to actually transact with a merchant's agent built on another's, without custom integration work every time (source: Axios, Forkast).
So what for brands: interoperability is the difference between agentic commerce becoming genuine infrastructure and becoming another decade of expensive point-to-point integrations that only the biggest players can afford. When you're briefing any vendor on an agentic build this half, the most useful question isn't what the tool does — it's whether it's built on these open protocols or on something proprietary. The proprietary answer may work today. It's the wrong answer for the medium term.

## What can Woolworths' stumble-then-guardrail approach teach every brand building an agent?

Woolworths is pushing both its digital assistants, Olive and Mandy, from answering customer questions to actually completing tasks — planning meals, reading handwritten recipes, filling shopping baskets. Olive already carries more than seventy per cent of contact-centre volume, so this is a rebuild of core infrastructure, not a side pilot. It follows an earlier, well-publicised moment where Olive wandered off script in customer chats, after which Woolworths reportedly built a layer of internal reviewers — informally, agentic judges — to check responses before customers see them (source: iTWire, Computer Weekly).
So what for brands: the most concrete, at-scale agentic-commerce deployment in the Australian market right now is as notable for its guardrails as for its features. The likely path for other large retailers isn't skipping the stumble Woolworths had; it's budgeting for their own version of it, and treating the review layer as a core product decision rather than a fix bolted on after an embarrassing transcript circulates internally.

## Is Southeast Asia about to leapfrog the US on agentic commerce?

Google is now building the AI shopping agents that will sit inside Shopee, giving Sea Group a direct answer to TikTok Shop and Lazada across the region — even as Sea has been trimming developer and QA roles globally to redirect spend toward AI. It lands on top of payment infrastructure that is already live: Mastercard's agentic payment rails are running in Singapore and Malaysia, backed by a regional AI Centre of Excellence, with Singapore having processed the first authenticated agentic transaction in Southeast Asia back in March (source: Forbes, Mastercard Newsroom).
So what for brands: this region isn't following the US playbook, where OpenAI's own instant checkout experiment stumbled on merchant onboarding earlier this year. The more plausible path in Southeast Asia is platform-level AI and payment rails arriving together, which is a faster route to agents actually completing purchases rather than just influencing discovery. Any brand selling through Shopee or Lazada should be checking, this week, whether its product data and pricing are structured in a way an agent can actually read and act on.

## The synthesis

Strip away the individual headlines and the pattern is consistent: the boring parts of agentic commerce — protocols, review layers, payment rails, buying interfaces — are the parts moving fastest right now, not the flashy demos. That's usually the more reliable signal that a shift is real. The one practical takeaway worth acting on this week is simple: pick one agentic surface your brand already touches, whether that's a media buying platform, a customer service bot, or a marketplace listing, and ask who owns the guardrails around it. If the honest answer is nobody yet, that's this week's job.

## Key takeaways

- The Trade Desk's Ask Koa signals that the media buying interface is becoming conversational; ask your agency how they plan to use it now, not after the case study.
- Google moving A2A governance to a neutral foundation is a step toward real agent interoperability; prefer vendors building on open protocols.
- Woolworths' Olive deployment shows guardrails and human review layers are core product decisions, not post-crisis fixes.
- Southeast Asia is combining platform AI and live payment rails faster than the US; check whether your Shopee and Lazada data is agent-readable.
- The practical move: pick one agentic surface you already operate, and name who owns its guardrails.

Source: https://marksuite.ai/insights/trade-desk-ask-koa-agentic-brands


## About the publisher

This analysis is published by MARK (https://marksuite.ai), a tech-enabled advisory service for AI model optimisation, AI search optimisation and agentic commerce optimisation. MARK measures how often AI models such as ChatGPT, Gemini, Claude and DeepSeek recommend a brand over its competitors, why they do it, and what has to change — across both content and the underlying data infrastructure (API connections, protocol integration, product feeds and structured data). Written by Malin Gaertig.

Related MARK pages: https://marksuite.ai/score (free Agentic Readiness Score), https://marksuite.ai/product (product suite), https://marksuite.ai/framework (framework guides).

Cite as: MARK — https://marksuite.ai