Chris Hillman

The Restaurant That Refused to Take Bookings OnlineLet me tell you a story about a restaurant owner who became obsessed with human connection.He didn’t want people booking online. He wanted them to call. He wanted the ritual of a human voice, the small exchange about an anniversary or a first date, the warmth of being recognised. His team thought he was losing his mind. Online bookings were standard. Everyone did it. Why make customers work harder?One of the managers finally asked him a pointed question: Have you ever actually made a reservation with us?He hadn’t. So he did.He called the restaurant. He was put on hold for thirty minutes. When someone finally answered, they were apologetic but firm — the restaurant was fully booked. No warmth. No conversation. Just a long wait and a closed door.In trying to humanise the process, he’d made it worse.Here’s where most stories would end with “so they moved everything online and fired the reservation staff.” That’s not what happened. They did move to online booking, but they kept the entire reservation team and repurposed them. These people now spent their days learning about the customers coming in that night. Who was celebrating a birthday? Who was on a first date? What had a regular not finished on their plate six months ago?The team became mini-concierges. Every guest walked in to find someone who knew them — not in a creepy, surveillance-state way, but in the way a good friend remembers what you’re going through. The technology handled the transactional layer so the humans could focus on the relational layer.I’m telling you this story because it captures something I think most organisations have forgotten in the rush to automate everything: the work of connection isn’t overhead. It’s the point. And as AI swallows the transactional layer of every industry at once, that relational layer is about to become the only point of difference left.Trust Is Built One Marble at a TimeBrené Brown tells a story about her daughter Ellen that I realate to. Ellen came home from third grade devastated. She’d shared something vulnerable with a friend at recess, and by the time class started again, half the school knew. She told her mum, through tears, that she would never trust anyone again.Brown’s first instinct, honestly, was to agree. Tell no one anything, ever. But she caught herself and offered a different idea — the marble jar.

There’s a jar in Ellen’s classroom where good choices add marbles and bad ones remove them. When the jar fills up, the class celebrates. Trust, Brown told her, works the same way. You share the hard things with people who have filled up your marble jar over time. Marble by marble. Small moment by small moment.Ellen immediately named two friends whose jars were full.What stopped Brown cold was what counted as marbles. One friend had scooted over at the lunch table to make “half a heinie seat” when Ellen had nowhere to sit. Another had remembered the names of Ellen’s grandparents at a soccer game. Brown couldn’t believe it. Surely trust required grander gestures than this.She went back to her research data and found exactly the opposite. The single highest-ranked trust-building behaviour in her entire dataset? People who attended funerals. Not the people who made big speeches or grand gestures. The people who showed up in small moments that mattered.Trust isn’t a transaction. You can’t purchase it, automate it, or accelerate it with a clever marketing campaign. It accumulates in tiny, unremarkable moments that look like nothing from the outside — and everything to the person on the receiving end.And here’s what I find most interesting about the marble jar: once it’s full, it’s incredibly hard to empty. You can weather bad days, miscommunications, even genuine mistakes. But if you’ve never bothered to fill it, a single bad interaction is all it takes. There’s nothing to cushion the fall.Every organisation has customers whose jars are full and customers whose jars are empty. The ones with full jars forgive outages, laugh off a late delivery, stay through a price increase. The ones with empty jars churn the first time anything goes wrong. Most organisations don’t realise which is which until it’s too late.Hospitality Is a Dialogue, Service Is a MonologueDanny Meyer, the restaurateur behind Union Square Hospitality Group and Shake Shack, draws a great distinction. Service is the technical delivery of a product — the food arrives hot, the bill is accurate, the room is clean. Hospitality is how the delivery of that product makes its recipient feel.Service is a monologue. You decide your standards and you deliver against them. Hospitality is a dialogue. You notice, you adjust, you respond.

Meyer puts it even more sharply: hospitality exists when you believe the other person is on your side. It’s present when something happens for you. It’s absent when something happens to you. Those two little prepositions contain the whole difference. You feel it instantly when it’s there. You feel it even more instantly when it’s not.Meyer isn’t anti-technology, by the way. But he draws a firm line: the technology he’s interested in is technology that reinforces hospitality, not technology that replaces it. He gives sommeliers Apple Watches for real-time information. He doesn’t give them to waiters, because waiters need to maintain eye contact with guests.Thats super important. A restaurateur whose restaurants have earned basically every award that exists decided his waiters should not be looking at screens because eye contact is more important than information. When was the last time you saw a decision like that made inside a large organisation?The best story I know in this vein is from Will Guidara, who ran Eleven Madison Park when it was named the best restaurant in the world. One night, Guidara overheard a table of European food tourists lamenting that they’d eaten at Per Se, Le Bernardin, and Daniel — but they were flying home the next day and had never tried a proper New York City hot dog.Guidara ran outside, bought a $2 hot dog from a street cart, convinced his chef to plate it with “swooshes of ketchup and a quenelle of relish,” and had it served as a surprise mid-course. Every person at the table said it was the highlight of not just the meal, but their entire trip to New York.A $2 hot dog in a room full of the most expensive food in America. That’s the gap between service and hospitality. You cannot design an algorithm that eavesdrops on dinner conversation and dispatches someone to buy a street hot dog, because the person on the receiving end would immediately sense the machinery of it. The magic is precisely that a human heard, a human decided, a human cared.Guidara was so moved by what that moment revealed that he created a dedicated role called the Dreamweaver — someone whose entire job was to make these moments happen. Researching guests ahead of visits, building cheat sheets so staff could greet people by name, listening for conversational cues during dinner. This is exactly what the restaurant in the opening did.

The technology handles the boring parts so the humans can do the thing only humans can do.The Things We Destroy Because We Can’t Measure ThemHere’s where I get frustrated, and this is the part of the article I’ve rewritten three times.I’ve spent a lot of time in and around banking over the past few years, and I’ve watched a particular kind of decision get made over and over. It always starts the same way. Someone pulls a report showing branch transaction volumes declining. Someone else pulls a report showing the cost per square metre of the branch network. A third person models what happens if you close the bottom quartile. The numbers are clean. The deck practically writes itself. And the decision gets made in a room where nobody has ever stood behind the counter and watched what actually happens on a Tuesday morning at half past ten.What you see when you actually stand there is not transactions. You see a retired schoolteacher who comes in every week not because she can’t use the app — she can — but because the teller knows her by name and asks about her grandson. You see a small business owner dropping off the weekly takings and mentioning, almost offhandedly, that he’s thinking about expanding into the next suburb, and the branch manager quietly filing that away because she knows a commercial lender who should probably call him. You see the elderly man whose wife just died, standing at the counter trying to change the joint account, and the staff member who recognises what’s actually happening and gently walks him through it for forty-five minutes without once looking at a clock.None of that shows up in the transaction volume report. None of it shows up in the cost-per-square-metre model. But it’s where the loyalty comes from. It’s the reason the schoolteacher’s kids bank with the same bank and her business-owner neighbour refinanced there and the grieving husband never even considered moving his accounts when the rate went up.When the branch closes, all of that evaporates. Not dramatically. Just quietly. The schoolteacher switches to the credit union down the road because they still have tellers. The business owner realises nobody at the new centralised lending hub remembers him, and starts taking meetings with competitors. The grieving husband’s children, watching their dad struggle with the phone banking menu, consolidate everything with a different bank when they inherit.

Three years later the spreadsheet shows a modest increase in customer churn and nobody connects it to the branch closure, because the causal chain is too long and the data was never collected in the first place.The broader pattern is consistent everywhere I look. Australian banks have closed something like 2,500 branches since 2017. The justification was always the same: digital adoption, customer preference, efficiency. But when whistleblowers from the Finance Sector Union testified at the Senate inquiry, a darker story emerged. Branch staff had been performance-managed to suppress in-branch activity — pushed to redirect customers to ATMs, pressured to sign them up for digital banking. The data that justified the closures had been partly manufactured by the closures themselves.And what the banks never properly measured was any of this. Every one of those moments is share of wallet. Every one of them is retention. And none of it appeared on the spreadsheet that closed the branch.The tragic irony is what came next. The same banks that couldn’t justify keeping branches open are now spending billions on AI personalisation engines designed to replicate the exact relationships they dismantled. JPMorgan Chase budgeted $18 billion for technology spending in 2025. Bank of America’s AI assistant has surpassed two billion client interactions. And yet only about a quarter of consumers say their bank provides tailored financial advice. They destroyed the organic version and are now paying orders of magnitude more trying to rebuild a synthetic one.There’s a concept I keep coming back to called the McNamara Fallacy, named after Robert McNamara, the US Defense Secretary who tried to run the Vietnam War like a business. McNamara was brilliant with metrics. A general once told him he needed to add an “x-factor” to his measurement list — the feelings of the rural Vietnamese people. McNamara wrote it down, asked what it meant, and then sarcastically told the general he couldn’t measure it, so he erased it.Daniel Yankelovich described the fallacy as a four-step descent. First, you measure whatever can be easily measured. Second, you disregard what can’t be easily measured. Third, you presume that what can’t be measured easily isn’t important. And finally — this is the fatal step — you presume that what can’t be easily measured doesn’t exist. Yankelovich called this last step “suicide.