AI Didn't Replace Retail. It Revealed It.
91% of NZ organisations are now using AI in some form - and only 4% say it's actually changed how they operate, down from 8% a year ago (Datacom).
Woolworths Australia reported that they ran 1500 GenAI use cases in a single year. And even by Forrester's own, more measured numbers, over a third of workers still fear AI will take their job within ten years.
Chaos. Scale. Fear. That's the landscape most of us are standing in right now, whether we like it or not.
But here is what gets lost the moment someone opens a laptop and starts talking about large language models and automation pipelines: this was never really a technology story. AI is a tool - a remarkable one, but still a tool. What it's being asked to serve is a business. And inside every business are people with judgement built over years, institutional knowledge that never made it into a spreadsheet, and customer relationships that exist because someone showed up, remembered a name, and solved a problem at 5pm on a Saturday.
Research is starting to confirm what the best retailers already know instinctively. McKinsey recently argued that the high-touch layer of retail - the service, the discretion, the pacing of a real human interaction - isn't a wrapper around the transaction. It's an essential part of it. And because it's becoming rarer, it's becoming more valuable. Human interaction matters even more to aspirational and younger shoppers, who are up to twice as likely to seek styling advice from staff compared with value-market shoppers. Not less. More.
The real question isn't how much can we automate. It's what does AI free my people up to do?
Four examples from the past six months answer that question in very different ways
The one where the backlash won
Coles - one of Australia's two dominant supermarket chains, and a direct competitor to the Woolworths Group that owns Woolworths here in New Zealand, spent this year finding out how much surveillance shoppers will tolerate.
In-aisle motion sensors built to catch organised retail crime went viral: boycott calls, privacy debates, accusations of overreach. Coles said it was motion-sensing, not facial recognition, and experts largely agreed. One shopper summed up the real issue better than any expert: ‘I'm not concerned about what the technology is being used for today. I'm concerned about what it could be used for tomorrow.’
Then came the bigger question: Coles' three-year data partnership with Palantir, the US analytics firm whose software also underpins American deportation raids and battlefield targeting systems. Coles maintained the deal covered rostering, supply chain and store operations, not surveillance. Customers weren't convinced.
Neither were investors: an 85,000-signature campaign, a shareholder resolution demanding a human rights impact assessment, and a regulatory finding that Coles had breached privacy law by failing to properly disclose the arrangement.
This month, Coles confirmed the contract won't be renewed past 2027. Nobody proved the technology itself did anything wrong. What proved costly was the silence around it. The lesson isn't ‘don't use AI for shrinkage or rostering.’ It's that trust, once spent, doesn't reset just because the technology turns out fine. Explain what you're deploying, and why, before someone else does the explaining for you.
The one where AI serves the human, not the other way round
Tecovas - ‘radical hospitality’ built into every store.
Tecovas builds handcrafted cowboy boots, direct-to-consumer, out of Austin, Texas. Their whole business model rests on something they call ‘radical hospitality’ - not a service layer bolted onto retail, but the operating philosophy itself. You're offered a drink. You're invited to slow down. Boot fitting is personal, considered and deliberately unhurried. It's the opposite of throughput retail, and it works.
Underneath that very human experience sits some genuinely smart AI. Tecovas now runs inventory allocation and replenishment across their entire store network, replacing spreadsheets and manual processes outright. The results: a 9.6% revenue lift in AI-managed categories versus human-managed ones, a 2% improvement in in-stock rates, new product sales up 20% through better initial placement, and allocator workload down 80%.
My favourite detail: they built an internal tool called Boot Runner in 36 hours using AI, so a store associate can request size inventory from the stockroom in seconds rather than minutes - meaning they stay present with the customer instead of disappearing out the back. The AI didn't replace the experience. It made the experience possible at scale.
The one that turns generalists into experts
Lowe's operates 1,700 stores, 300,000 staff and 35,000 to 40,000 SKUs per store, serving 16 million customer transactions a week. Most associates aren't tradespeople. A customer walks in needing a plumbing fix; the person in the aisle sells paint. Familiar problem, for anyone running a large-format store.
Lowe's answer is MyLow Companion, built into the handheld devices staff already carry: real-time inventory, exact product location, project-based recommendations and - because associates are moving, carrying stock, one hand on a trolley - voice-activated and hands-free.
MyLow Companion puts real-time answers in the hands of every associate, tradesperson or not.
Over half of all interactions happen by voice. A million questions a month across the customer and staff-facing versions combined. What used to take three people thirty minutes now takes one associate a few.
AI didn't replace the sales assistant. It made the associate the expert they didn't previously have the training, or the time, to become on their own.
The one that shows what's already normal elsewhere
I saw this firsthand just a week ago, leading a delegation of Australian retail leaders through Shanghai and Shenzhen. At Tencent, we watched their workplace platform Lark generate a video from a single photo and one sentence of instruction. Minutes later: a finished, unnervingly real clip. No production crew, no edit suite, no approval cycle.
Then there's WeChat, which doesn't feel like an app so much as connective tissue - payments, messaging, ordering, loyalty and identity, all invisible until the moment you need it. It isn't that China has better technology. It's that they stopped treating each function as its own product years ago.
The lesson isn't ‘go build a super app.’ It's that the retailers moving fastest right now have stopped bolting AI onto existing processes and started rebuilding around what's actually possible. That's a much bigger gap to close than most boardrooms have budgeted for.
Where this is all heading, if we let it
Here's the future the biggest platforms are building toward, whether retailers ask for it or not. Shopify has already wired in agentic checkout for ChatGPT, Google's AI Mode and Gemini, and Microsoft Copilot.
Picture a shopper telling their AI agent to buy a bag when the price drops from $150 to $90. The agent monitors hundreds of retailers, waits for the trigger, and buys. No browsing. No discovery. No relationship with the brand at all.
That's not a future I want retailers operating in. Real retail isn't just a transaction - it's discovery, desire, a customer falling in love with something they didn't know they needed and paying full price because the experience made it worth it. That customer is worth ten AI-optimised, price-triggered purchases.
Coles shows what happens when trust runs out. Tecovas, Lowe's and Tencent show what it looks like when AI earns its keep. The retailers who get this right won't be the ones who handed everything to the machine, or hoped nobody would notice.
They'll be the ones who gave their people room to do what they were always good at, and told customers the truth about it.
Comments