even asked. And it needs your camera roll too. If you’re anything like me, that holds a photo of Tuesday’s whiteboard session sitting between last night’s dinner and a picture of your kids. Whenever privacy comes up, someone always says they’ve got nothing to hide. But I’ve never thought that was the point. Privacy has very little to do with hiding. It’s the right to choose who sees which version of you, and when. You’re a different person in a company meeting than you are waiting at the school gate for pick-ups. That’s completely normal and human. Article 12 of the Universal Declaration of Human Rights enshrines that right, and it has nothing to do with whether you’ve done anything wrong. Australians seem to sense where this is heading. Ping Identity’s 2026 consumer survey found that only 18 per cent of Australians are open to letting AI make decisions on its own, compared with 70 per cent in India. Meanwhile, the OAIC’s 2026 Australian Community Attitudes to Privacy Survey found that only 4 per cent of respondents trusted AI companies, and 3 per cent trusted social media companies. Yet despite this resistance, we already share everything. It doesn’t make sense, right? I see these contradictions all the time. I have friends who are completely against giving away their personal information, yet share their lives on social media every single day, using an iPhone that’s gathering everything they do. I’ve also raised this in talks and workshops I’ve given. I’ll ask the room what people think of Meta and similar corporations, then ask how many use Instagram or TikTok, even though those platforms are collecting information about them every day. And if I’m honest, I’m no different. The more I understand how AI works, the more I realise I’m giving away far more access to my data than I realise. This contradiction has a name: the privacy paradox. But the more you look at it, the less it looks like a paradox. People aren’t necessarily inconsistent about privacy. They’re simply consistent about two other very important things. The first is control. When you post a photo of your kid on social media, you choose the photo, write the caption and decide roughly who sees it. It feels like an act of authorship and choice. In 2013, researchers at Carnegie Mellon found that giving people more control over what got published led them to disclose even more sensitive information. Yet the real risk lay in who could access that information afterwards, something they couldn’t control. The second is context. Helen Nissenbaum’s work on privacy shows that people judge information by whether it’s flowing as expected within a particular setting. A photo of your child shared with your friends fits the norms of that relationship. But using the same photo to train a model breaks those norms. So the objection was never really to sharing itself. It’s about information leaving the room it was meant for. An agent takes both of those away at the same time. There’s no moment of decision. One permission can cover a thousand actions you’ll never even see. Information moves between different parts of your life without you necessarily noticing, because that’s exactly what an agent is designed to do. Suddenly, the whiteboard, the bank account and your colleague’s calendar all feed into the same booking. You can see exactly where people start to hesitate. Accenture’s 2026 consumer research found that 74 per cent of people would trust a personal AI agent more than their best friend to make a purchase for them. That sounds like a green light until you read further. Only 32 per cent would let an agent choose within set budget and brand limits, 9 per cent are open to fully autonomous purchasing, and just 12 per cent would allow it to handle payment. People want the help, but recoil at the handover. And that’s precisely when the agent needs access to their information. The industry is asking for trust it hasn’t earned yet Every worthwhile relationship you’ve ever had has worked the other way around. Someone does a small thing well, then does it again, and you give them a little more. That repetition builds reputation, and reputation earns the next yes. That’s how trust has always worked, and it won’t change. But agentic commerce is being rolled out backwards. It asks for everything on day one and promises trust will follow later. A brand that wants your bank login before it’s even booked you a single flight hasn’t proved anything yet. We’ve already seen what it looks like when a brand gets this right. When I was at Nike, we launched Nike Run Club. It asks for quite a lot of information from the day you join, including where you run and how fast. Yet runners handed it over without thinking twice because every run gave them something back: their pace, their route and a clear picture of how they were improving. They could see and feel the value from the moment they stopped. The access matched the job, and the value showed up every time they laced up. Trust me, that repetition is exactly what earns the next permission. That’s the type of trade people make with their eyes open. This gets really uncomfortable when it comes to consent. The OAIC found 69 per cent of Australians always or often accept privacy policies without reading most of them. This group is also the most likely to feel consent is rarely a real choice. Let’s be clear: there’s nothing wrong with a brand nudging a customer towards something that genuinely solves their problem. I’m all for that. It’s exactly what a good agent should do. But leaning on someone’s resignation to secure their consent is different. An “accept everything or don’t use it” approach does exactly that. Consent given out of resignation isn’t trust, and it’ll break the first time something goes wrong. Most people don’t read the terms because they’ve already decided refusing won’t change anything. When a product is designed around that resignation, the yes it collects tells you nothing about whether the customer trusts you. It only tells you they felt they had no choice. Those are two very different numbers to be running a business on. Then there are the people who never even got a say. The child in the photo didn’t consent, and neither did the colleague in your calendar. Our convenience is being built partly on their information too. I already know the reply to all of this. We said the same about smartphones and social media, and then convenience won anyway. But there’s an important difference. Previously, information tended to stay within individual apps. An agent only works by joining it all together. That makes this a much bigger trade than the one we made before, even if it feels familiar. An agent can access your photos, your bank, your passwords and everyone in your calendar all at once. With all that information connected, one mistake doesn’t necessarily stay contained to one app. And the fact that convenience won last time isn’t proof we got that trade right. It might just be proof we never really got asked. Retailers are about to be trusted second It’s easy enough to read all of this as a fight between consumers and big tech and leave retail out of it. But I think retailers have the most to lose here. When an agent sits between your customer and your brand, the customer trusts the agent first. You’re left borrowing whatever trust remains. If customers won’t grant access to their information, agentic commerce stalls long before it reaches your checkout. And if they do grant access but something goes wrong, they won’t remember which part of the chain failed. They’ll remember where they were shopping when it happened. Retail should know better than most industries how trust is built: one kept promise at a time, often when it would’ve been easier not to keep it. That’s the advantage here. Brands that apply that same principle to how they ask for access, rather than handing the whole relationship to whichever platform asks first, are the ones customers will let in further. Ask for less, earn the rest Here are a few things worth testing as soon as next week if you’re building towards agentic commerce or partnering with someone who is: For every permission you ask for, name the moment the customer will see it working for them. If you can’t, you haven’t earned it yet. Start with the smallest version of the job and make the next permission something the customer chooses after watching you do the first one well. Ask who finds out first when something goes wrong: you or the customer? If it’s the customer, you’ve built your convenience on borrowed trust. So is convenience worth your privacy? I think it can be, but only if we decide deliberately, together, as a community, rather than letting defaults decide for us. Those defaults are being set right now by the people who profit most from a yes answer. Right now, the race is about capability: who can do the most on the customer’s behalf. I don’t think capability wins this one. The agent people actually let in will be the one who asked for the least – and earned the rest. Nick Gray is the founder and CEO of IGU Global, a Sydney-based retail and brand strategy consultancy.