What AI Should Never Be Allowed to Do Without Us

Artificial intelligence is disrupting the world faster than any previous technology change ever. The next decade of retail will be won by the brands that use technology to deepen human relationships, not replace them.

A few weeks ago I stood at the back of a London flagship and watched a sales associate greet a returning customer by name, ask quietly about her daughter’s wedding, then bring out a piece she had set aside the month before. It took 90 seconds. The customer made a five-figure purchase. That moment was not the product of an algorithm. It was the product of memory, the single most undervalued asset in retail.

I've spent my entire career in retail, much of it inside or alongside luxury and premium brands. I've watched the pendulum swing from store-led to digital-first to omnichannel, and now to whatever we're calling this current moment of agentic AI and autonomous customer journeys. Every wave has promised to transform the customer relationship. Most have transformed only the org chart.

The brands that will define excellent retailing will not be the ones with the largest data lake or the boldest tech announcement. They will be the ones that use technology to give their people the time, the context and the permission to do what only people can do: notice things, remember things, and make a customer feel known and valued. Technology can amplify the customer relationship, but a human is always at the heart of it.

The Paradox at the Heart of Retail

Here is a number that should rattle every retail executive: in a 2024 IBM global consumer study of 20,000 shoppers across 26 countries, nearly 75 percent said they prefer to shop in stores, yet fewer than one in 10 were satisfied with the experience they actually find there.

Read that twice. The customer wants the store. We're failing to give them one worth visiting.

And it's not for lack of investment. Over the past decade, retailers have poured extraordinary sums into clienteling platforms, virtual try-on, RFID, livestream commerce, NFT-gated drops, and more recently, generative AI. LVMH’s MaIA agent is now handling more than 1.5 million internal requests a month. Richemont has connected 11 maisons across 25 countries to a single AI-driven recommendation engine. Sephora has plugged 45 million Beauty Insiders directly into ChatGPT.

The technology works. The customer relationship, however, in many cases is still broken.

That is the paradox retail has to confront. We've never had more data about our customers, and yet customers have never been less certain that they feel known by us when they walk through the door.

The Pendulum Has Swung Back, But Smarter

For nearly two decades, the dominant story in retail was that physical stores were a problem to be optimized away. That story has quietly collapsed. E-commerce as a total share of retail in the U.S. has plateaued at around 16 percent. The direct-to-consumer darlings of the 2010s are quietly opening stores or being sold for parts: Allbirds, once valued at $4 billion, recently changed hands for $39 million.

In luxury retail, the picture is even sharper. Customers have always wanted the store. They want the ritual, the welcome, and the selling ceremony. What they will no longer tolerate is a store that treats them as a transaction.

The most interesting retailers I work with have stopped arguing about online vs. offline and started asking a more useful question: What is each store actually for? For some, it's a stage. For others, a salon. For others still, an omnichannel hub. The brands losing share are the ones still treating every door the same.

Why 'Agentic AI' is the Wrong Starting Point

On April 16, at its Capital Markets Day in Florence, Kering told investors it was rebuilding its client infrastructure around “agentic AI” — autonomous systems that can schedule follow-ups, adjust recommendations and trigger outreach without a human pressing "Send." A new chief digital, AI and IT officer now sits on the executive committee. CRM has been promoted out of marketing into its own division called Clients.

I admire the ambition, yet I'm wary of the framing. The question retail leaders should be asking is not what can AI do on our behalf, but what AI should never be allowed to do without us.

In mass retail, an autonomous agent firing off a re-engagement email to a lapsed customer is a productivity win. In luxury, the same message, sent at the wrong moment, in the wrong tone is a relationship ended. Luxury sells intimacy and experience as much as product.

The next generation of CRM leaders in retail will not be measured on campaign open rates. They will be measured on something harder: whether they can set the rules, the constraints and the editorial judgement within which AI is allowed to act on behalf of a maison.

3 Things Luxury Brands Should Stop Outsourcing

If the technology debate has been asking the wrong question, the right one is more grounded: Where does the brand show up, in person, and how do we make sure that moment is unmistakable?

1. Memory: Clienteling tools are often sold as data-capture engines. They should be sold as institutional memory. The point of knowing a customer’s preferences is not to feed a segmentation model. It is so the next advisor can continue the relationship naturally.
2. Welcome: In luxury, the first 30 seconds at the threshold of the store is where loyalty is decided. Not the website. Not email. The threshold. Technology should always be in service of the welcome, not the other way around.
3. Restraint: Luxury is defined by what a brand chooses not to do. Not every customer wants a WhatsApp message. Not every product needs a personalized lookbook. Not every interaction should be optimized or automated.

Some of the most valuable signals luxury brands send are restraint and patience. AI is uncomfortable with both.

What to Actually Do on Monday Morning

For the executives I speak with each week, the practical question is rarely “Should we deploy AI in CRM?” It's already being deployed, often in fragments and without a clear owner.

The harder questions are these:

• Who in the organization owns the end-to-end relationship? If ownership is split across teams, the customer ends up paying the cost of your org chart.
• Are your store associates truly empowered or is technology just deployed? AI now makes personalized training cheaper than ever. The challenge is no longer access to technology. It is whether teams are trusted to use it well.
• What are you measuring? Luxury brands still over-index on transactional metrics. Relationship metrics matter more.

The Future is Human

Luxury has never sold things. It has sold a feeling of having been chosen, recognized, anticipated, and quietly looked after. Every technology wave will keep promising to scale that feeling. None of them ever fully will.

The work of luxury, in 2026 as in 1926, is still done at the store threshold, by a person who has been given the time to pay attention.

The brands that remember this will own the next decade. The rest will keep adding disparate tools and apps, wondering why their customers no longer feel like guests.

The future is still very much human.

Cathy McCabe is the CEO and co-founder of Proximity Insight, a retail technology company focused on helping brands deliver more personalized, data-driven customer experiences across channels.