Were you forwarded this email? Sign up here to get the next edition.

I was in Sydney this week for the launch of Thread Together's Fashion Impact Collective. It's an incredible initiative. More than 70 brands have already signed up, including The Iconic and David Jones. The aim is to clothe people in need, keep excess stock out of landfill and support one of the most important charities in Australia.

If you're a fashion brand, it costs between $500 and $3,000 to take part for the year. Let me know if you'd like to connect and be one of the inaugural members. Bonus... it has nothing to do with AI. Unlike this week's newsletter.

CHEERS, BUSHY

01

If I see one more webinar titled "Is your store agentic ready?" I'm… well… I'm not going to go. But Bill Hunt ran a test on 122 Shopify stores for Search Engine Journal this week that's worth a read. Good news first. If you are on Shopify, you're agent-ready whether you asked for it or not. Shopify switched its WebMCP tools on for every Liquid storefront back in August. All 74 live checkouts behaved exactly as Shopify designed, including stopping for human confirmation before payment.

Now the gap. 83 of the 89 stores with an agents.md file copied Shopify's template and never customised it. And the agents struggled with basic buyer questions. None of the 88 stores tested could quote shipping to a specific city. Only 2 could explain their cancellation window. Returns were better but still patchy. There's a really good list of actions at the end to make sure agents can read your rules before Black Friday. Start with your agents.md and your policy pages.

Potentially the most heartwarming story this week shows our robot friends have some heart. A Reddit user asked ChatGPT to buy them a mystery gift on Amazon under £20. Something that would make them laugh, based on everything it knew about them. It bought them a book called How to Invent Everything: Rebuild All of Civilization (with 96% fewer catastrophes this time). Sure. Robot humour. We are truly fucked.

But… it's not as creepy as what Amazon is already telling people it knows about them. From "flat bottom buttocks" to "cusses when scared" to "cannot keep real plants alive", it's not even trying to hide it. You can see how Amazon is pigeonholing you under Account > Shopping Preferences > About You. It doesn't look like it's live in Australia yet. But it's there for everyone in the US to see.

02

While we're on agentic, Ping Identity surveyed 1,000 Australians about trusting AI. We'd let an AI agent spend an average of $200 on a routine purchase without final approval. But more than half would cap it at $143. Who said we weren't a nation of gamblers? Globally, only 3% are comfortable with bots transacting for them. I don't know about you but I'm seeing a lot of people take out credit cards with very small limits to start experimenting with agentic commerce without much risk attached. I blame Gen X discovering Meta Muse for this.

To get ahead of the agentic game, American Express has introduced an Intent Intelligence Framework. The customer signs off on what they want before the agent shops. The merchant, the budget, the product. If you asked for white Nikes and the agent tries to buy Adidas, Amex can stop the payment. We've got bots checking bots' work to keep us safe. I smell purple fire.

But the people I really want to meet are those who took Elon's advice, pressed the chaos button and sent Grok a photo of their credit card so it could shop for them. It's not an agentic Christmas this year. But I reckon next year we're in for some fireworks.

03

Let's keep going with the good news, hey. Nozomi Networks Labs found Australia is now the most targeted country for IoT and OT cyberattacks, up from third last time. Thankfully retail didn't make the top three sectors (ie. we have no money). The most active group was Scattered Spider, an organised crew that relies on human trust to get into systems.

ASOS was hacked last week after attackers impersonated a trusted contact to get into an employee's account. They got customer names, contact details and search histories. Its shares dived more than 10%. I know those corporate phishing tests are really annoying. They've never been more important.

Human error is the main cause of security breaches in ecommerce. And we have to be careful even with third-party apps approved by Shopify, because they can be infiltrated too. Swym Wishlist was hit recently. I spoke to the team behind it. They said it was resolved within two hours and extra security measures were in place within 48 hours. Lessons for everyone.

But we've got some white AI knights on white AI horses. Sydney startup Apate.ai creates AI versions of scam victims to bait scammers and gather intel on what they're running, without anyone getting hurt. Its agents have already wasted around 17 years of scammers' time across 2.5 million conversations. These are the productivity gains we've all been searching for as a human race.

04

But don't be afraid. AI isn't the end of the world. Influencers still are. Tesco is bringing influencer content to more than 3,800 digital screens in its UK stores. Shoppers will get their influencers served up right in front of them as they shop. Can we get Albo onto it? Surely it's harmful for our children to see influencers on big screens in public.

And while we're talking about our children's future, Square has launched Square Kiosk. It's an interactive, voice-enabled kiosk built for quick-service restaurants (US only for now). It's obviously aimed at sending more 15-year-olds out into the street. I was just getting used to the randomness of opening the Maccas bag at home to see what they ended up putting in there.

05

After a couple of years of taking away advertisers' control over where their ads appear, Google and Meta are now making a bit of noise about the control they're giving back. Google's new channel prioritisation lets you set channels to high or low priority across Search, Shopping, PMax and Demand Gen, and cap what the low ones get. Apparently we're lucky enough to weight our investment towards where we want to show. No promises though. Google calls its caps "best effort". Meta has gone the other way. Sales and leads campaigns can no longer pick placements at all. You just get to nudge bids up or down by placement. I feel so in control.

Over at Bing Webmaster Tools (which I know you're in every day), Microsoft has added a Topics Portfolio to its AI performance reporting. It shows the topics your brand is showing up for in AI answers, plus the queries where other sites get cited and you don't. You need enough Bing traffic for it to appear. But Microsoft is really delivering when it comes to free insights at the moment.

ATC PLAYBOOK #673

Your AI ads don't look like your brand. That's rarely the AI's fault.

Most ecommerce teams have now had a go at making ads with AI. A good number have stopped. Not because the output was wrong exactly. It just didn't look like their brand, and the moment they fixed one thing, everything else went a little wonky.

Sam Kroonenburg built A Cloud Guru with his brother and sold it for just over $2 billion. Then he built Cuttable, an AI ad platform that now holds ad-by-ad data on more than $320 million of Meta spend. Before building it his team sat down with 50 to 70 businesses to work out what was going wrong. In today's Playbook we cover the three things you control before any tool generates a thing. The material you feed it. The rules you write before you hand over the keys. And clear direction on what your brand actually means.

Examples from this Playbook

✂️ Sam from Cuttable found brands sitting on Drives full of paid-for shoot assets they couldn't afford to recut. The more control his customers get over each layer of an ad, the more they publish.

📱 Sam from PYRA shoots every look on his iPhone alongside the pro crew. One of his best top-of-funnel ads worldwide is an iPhone clip of a guy standing by a river. It's been running for two years.

🐳 Zach from Triple Whale shared how True Classic lets an AI agent run its entire Meta budget. But only after writing a ten-page file of every decision a human would make.

🧬 Kate from The Memo built a brand DNA document the whole business still runs on. Influencers, offers, hires. Every decision goes back to the DNA.