Category Archives: work

Oct 09 2026
Leave a comment
work

The Art of Working Within Bounds

I have written before about Jamie Fobert’s extension to Tate St Ives, precisely because there is something about the way the project dealt with constraint that has resonated.

I should be careful here. I know enough about architecture to be interested in it, certainly not enough to tell architects how to do their jobs. But one part of the Tate project felt to me very familiar from my own work. The building had to accommodate large artworks arriving through tight Cornish streets, then somehow move them through a site embedded in a granite cliff. These were not details to solve once someone had designed the pretty bit. They helped determine what the building could be.

I was reminded of it again by something rather less glamorous: a footbridge over the railway in Wokingham.

Tan House Crossing is the sort of infrastructure most of us would ordinarily walk across without thinking much about it. Network Rail needed a bridge that could accommodate the railway beneath it, including future electrification, meet the safety requirements of a modern rail crossing and, rather wonderfully, actually get to Wokingham in the first place.

That last constraint has consequences. Large sections of the bridge were to arrive by rail, but the route to the site passed beneath another low, narrow bridge. So before anyone became too attached to a particular structural idea or a handsome form, there was a fairly blunt physical limit on how wide the thing could be.

Then there was the steel. Rather than specify a structure and order new material to suit it, the team wanted to use repurposed steel stock, because carbon. Digital modelling allowed them to adjust the structure around the sections that were actually available.

I like this because it reverses a habit that is just as common in digital work. We tend to imagine the ideal thing first, then discover the organisation, technology, regulation, budget or operating model cannot support it. At which point everything becomes a negotiation about how much of the original idea can survive. Tan House started with what was true.

That did not mean accepting a miserable result. A narrow bridge with tall protective sides could easily have become a slightly oppressive steel corridor. Instead, the structure was worked down into the sides and the parapets splay outwards, giving people more space and preserving the view as they cross. The engineering requirement is still there. You just experience rather less of it. It’s this that feels closest to the work I now do through Senvinter because lot of experience problems arrive thinly disguised as requests for a better interface, lipstick if you will on the proverbial. Redesign the search. Simplify the finance journey. Improve the checkout. Fix the navigation. Often I find a more useful first step is to map the things the interface cannot simply wish away: regulation, product rules, legacy technology, fulfilment, data, organisational ownership, decisions made elsewhere in another time.

Such constraints are not necessarily the enemy of the experience. They are the material you one has to design with.

In financial services I spent years working around rules governing what can be said, recommended or disclosed. In automotive too there are finance structures, market differences, stock systems and legal requirements sitting behind what, to the customer, ought to feel like buying a car. You can ignore those things while drawing the ideal journey, but they will still be waiting for you when someone actually tries to build it. Increasingly I think the valuable work happens somewhat earlier. Establish what is fixed. Find out what only appears to be fixed. Work out which constraint is specifically causing the problem and which ones are merely inconvenient. Then make choices around reality rather than treating reality as the disappointing phase after the workshop.

There is some genuine craft in that. Tan House Crossing is interesting precisely because the constraints are still so legible in the finished artefact. The available steel affected its structure. The railway affected its dimensions. Safety affected its sides. Yet somebody in the process still paid attention to what it would feel like to walk across.

Look, I am wary of extracting grand universal principles from a footbridge in Berkshire, but I do think there is something useful in that sequence. Understand the bounds first, then design.

AI: AI found the Tan House Crossing example as part of an occasional trawl for interesting stories aligned to the way I think. I then wrote the piece.

Tagged , , , , , , , , , , ,
Oct 07 2026
work

The Cookbook Was Never the Problem

There is nothing modern about this problem. It is one that smells of dust, glue, vanilla sugar, and echoes with the defeated sigh of domestic ambition. It lives in the bowed cookbook shelf, in those cracked and fraying spines, in the turmeric fingerprints on books bought under the same repeated fraud: this one will be used.

The kitchen has always been an archive of sorts, albeit one with little hush and a source of fire, collecting fat, heat, paper, steam, and intention. In this setting the home cook becomes an unpaid librarian of appetite, hoarding recipes with the same touching optimism previous generations reserved for jam jars, opaque repurposed ice cream tubs, bits of string, and screws, so many screws, in tins. Our collections look magnificent: a decade old hardback with a television grin on the cover, a recipe pamphlet from a package holiday ‘pasta night experience’ stapled through the heart, a tatty box of clippings, dead URLs, orphaned instructions. Dense, generous, full of possible suppers. Then the evening’s plan is discussed. A surfeit of soon-to-be-mush courgettes in obscene quantities. Chicken again, though not like Tuesday. Anything that is not another traybake committed to the oven and considered an answer. At that moment that bowed shelf turns from abundance into accusation. You do not lack recipes. You lack retrieval.

Cookbooks are almost perfect, as anyone who’s last-minute panicked in Waterstones on Christmas Eve can post-rationalise. They do not ping, dim, update, lock, refresh, or retreat into a black mirror when your hands are slick with garlic and chicken juice. They sit open, obedient and useful, inspiring. They are tools. A phone is never just a recipe. It is a siren of distraction: messages, adverts, pop-ups, consent banners, and some foodie stranger’s 1,500-word emotional journey with parmesan. Consequently, we must resist and insist the answer is not another app-shaped panic, not any kind of system, not a deluded Sunday afternoon of organisational typing-it-all-out righteousness.

Go to your shelf, have an app produce an index. Barcode, title, author, tags: searchable, private, useful. Rescue them from their own splendid inert stupidity. So when those courgettes arrive in vulgar green abundance, you consult your own shelves first. Not that shouting bazaar of the internet, nor the monetised blog-ramble of strangers. Your library, the one acquired through singledom, coupledom, a Tuscan period, health kicks and “the kids will eat like lords!”. Your taste. Your choices, made earlier, when you were thinking clearly or when others were thinking of you. Search. Find. Pull the book. Cook.

It was never an absurdity owning too many cookbooks, the absurdity was standing in front of hundreds of pages of tested, trusted, tactile ad-free intelligence and choosing instead to prod at a sticky, dimming, impatient rectangle to find out how long to roast a beetroot.

AI: I used a bit of ChatGPT as a sub-editor.

Tagged , , , , , , ,
Sep 29 2026
work

BMW Neue Klasse and a better way to think about warnings

There’s a lovely small detail in BMW’s Neue Klasse safety engineering that interested me rather more than the breathlessly inevitable claims in the press pack about computing power.

If the car detects a cyclist approaching while you start to open the door, it can briefly stop the electronic lock releasing. You pull the handle. Nothing happens for a second or so. Cyclist passes. Door opens.

It is in itself a tiny intervention, but it tells a story about the way we have designed automotive interfaces for years. The usual approach would be to detect the cyclist, flash something, perhaps in the mirror, make a noise and hope the person opening the door notices the warning, understands it and reacts before depositing said cyclist across the road, GoPro and all.

We have become very good at warnings. Cars bong, flash, vibrate and illuminate little amber hieroglyphics all over the place. Much of the time this is eminently sensible. But every warning also represents a decision about responsibility: the system knows something, tells the dumb distracted human, then leaves the same fallible human to sort it out.

In this case BMW appears to have asked a more pertinent question. Given that the car already knows there is an immediate hazard, and given that delaying the door by perhaps a second has almost no meaningful cost, why involve the human in that particular bit of the decision at all?

We spent quite a lot of time thinking about this during the Human–AI Systems course I took this year with Cambridge University. Discussions about automation and Ai tend to drift towards capability: what can the machine do, how clever is it, how much should we automate? The problem is usually much more prosaic. Who is actually best placed to do each part of the job?

Here, the person still decides to open the door. The car does not second-guess where they are going or decide they should remain imprisoned in a BMW for their own good. It simply refuses to execute that intention for the brief period in which doing so would be obviously dangerous. That feels like good judgement.

And it connects to something I like finding in much broader experience work. When a problem becomes visible at an interface, the natural response is to improve the interface. Make the warning clearer. Increase the contrast. Rewrite the copy. Add another confirmation step. It’s catnip on social media where junior designers are endlessly posting ‘better’ forms and processes.

Sometimes all of that work is downstream of the more useful question: why have we made this the user’s problem?

A better warning cannot compensate for badly allocated responsibility.

There will obviously be limits to this. Systems that intervene too readily become infuriating, unpredictable and eventually distrusted, we can all think of examples where this is the case. But this particular example is wonderfully clean: immediate hazard, high confidence, trivial delay, potentially nasty consequence.

So yes, BMW’s new computing architecture is probably very clever but not as pleasing as the fact that someone used a little of it to make one warning unnecessary.

AI: Ai was used to correct a smattering of the most egregious grammatical errors.

Tagged , , , , , , , , ,
Sep 25 2026
work, cycling, socio-cultural

Cycling’s PizzaExpress Problem

You never really pay full price at PizzaExpress. When I was nineteen it felt like a nice meal out. Now someone checks for a voucher before anyone orders, and paying the price printed on the menu feels reckless, incompetent. The pizza may thankfully be much as it was. What has changed is what customers think it ought to cost.

Le Col, late of this parish, had a version of this problem. Complete a Strava challenge, get money off. A challenge this summer offered 30% off full-price kit. It was clever: cyclists were already on Strava, already inclined to chase a target, and the reward sent them to the shop. It also gave them a reason to wait. Once you have bought a jersey with a code, the price without one becomes a penalty for being impatient.

Personally I never really bought into Le Col, although I could see its appeal. Yanto Barker had raced professionally, the kit looked good, and you could believe that someone who had spent years in the saddle might have useful opinions about where to put a seam. Its former operating company entered administration in June. The offers alone cannot explain that, but they leave whoever carries the name forward with an awkward job: persuading riders that the price on the label is a real one.

Rapha is different for me. I remember what it felt like when it arrived on the scene. It made road cycling and courier-style urban commuting look stylish while keeping the suffering, odd rituals and long history intact. The clothes, photography and writing belonged together. For a British company to take that idea across Europe, Grand Tours and into the US was quite something. Rapha helped establish the visual language that newer brands now work within.

Today its website still calls the clothes “the world’s finest”. That is an extraordinary thing to ask a customer to continue to accept. There was a time when I might have let the boast pass on design grounds alone. Rapha made some of the kit everyone else seemed to be reacting to. Now many of us would look through the range and struggle to see what earns a claim of superiority across all of it, perhaps in some areas of technology and some of design, but universally?

The UK men’s shop alone currently shows 467 product listings. Alongside jerseys and bib shorts are shoes, sunglasses, hoodies, bikepacking bags, bar tape and a bookstore. There is nothing inherently wrong with a broad range, and the number includes many perfectly sensible products for different sorts of riding. But breadth changes the nature of the claim. “The finest” sounds rather different when it has to stretch from a racing skinsuit to a pair of casual trousers.

Rapha’s own account of itself still sounds like the brand I admired: demanding rides, clothes tested by professionals, stories told through photography and film, a club that at least says it “makes cycling sociable” though is almost certainly a fierce clique of try hards Much of that work is real. But I don’t think a professional rider testing one collection settles the quality of everything in the shop, or that a well-made film tells me which of several near identical jerseys is worth its price. The story has become larger and more confident while my reasons for believing every part of it have become less clear.

It is also a far more familiar sight on the road. That is success of a type, of course. Rapha opened up a particular idea of cycling to a lot of people, myself included. But a brand that once made you feel you had found something can’t depend on that feeling indefinitely. The product has to carry more of the weight once the badge is commonplace and demographically diverse.

MAAP and Pas Normal Studios now occupy some of the space Rapha once owned. They run offers too, so this isn’t a claim that their pricing is pure or every garment better. Their advantage, to my eye, is that each has a more recognisable point of view. MAAP can be loud and restless; Pas Normal can look like the entire club has agreed on a dress code without telling anyone else, a cycling murmuration. I don’t admire every choice, but I can usually see the choice being made.

Rapha says it is reducing its reliance on promotions. That is a difficult thing to do when customers have learnt the sale calendar, and its latest financial figures offer little comfort yet. Fran Millar also set about reducing the number of product launches and putting more attention back on the kit. She has since stepped down as chief executive, with further cuts planned. Unlike the ambulance chasers on LinkedIn, I wouldn’t use the accounts to declare her approach a failure. Recovering those full-price customers takes longer than ending a succession of sales.

Le Col shows the pricing problem in its most obvious form: too many opportunities to buy at a discount can make the advertised price hard to believe. Rapha has a more daunting task. It has to decide which parts of a large, successful brand still justify the claims made on its behalf, and which are there because a large business kept telling its team to keep finding new things it could sell.

On the Rapha homepage today, the invitation to shop the new collection sits alongside an offer to bundle and save up to 20%. I can understand both messages. I’m less sure they belong to the same idea of “the world’s finest”.

AI: Ai naturally helped me get some wider context on what’s been going on and checking some details. I wrote the piece.

Tagged , , , , , , , , ,
Sep 17 2026
work

Would your future self still buy it?

As is perhaps par for the course in ‘think pieces’, this one starts with the fact I came across a study recently that got me thinking about what ostensibly is a fairly simple question.

Think back to the person you were five years ago. Now imagine yourself five years from today. Which of those two people feels more like you?

My instinct was that the person in the past would feel closer. After all, I know him. I can remember what he did, what sort of things he cared about and some of the decisions he made. It is concrete, or at least salient. The future version is, by definition, imaginary.

Apparently, most of us see it the other way round.

The study published this summer in Personality and Social Psychology Bulletin looked at this across 12 studies involving more than 11,000 people in nine countries. Participants tended to see their future selves as more similar to who they are today than equally distant versions of themselves in the past.

Can you see the nice little trap in that? We can see how much we have changed because we have the evidence: jobs we once thought were perfect, clothes we wore, things we spent money on, opinions we would now rather prefer not to have quoted back to us. But when we look forwards, we tend to start with the person we are today and extend the line out, it is a line of betterment and the past shows the iterations from a more inferior you.

The researchers found something else too. People expected their future selves to be happier with decisions made today than they themselves were with equivalent decisions made by their past selves.

So I can look at a decision made by 2016 me and think, “Well, I wouldn’t do that now”, while remaining pretty confident that 2036 me will appreciate the excellent judgement displayed by 2026 me.

And, happily for someone in my industry and sphere of work, there’s probably quite a lot of consumer behaviour tucked up in there.

Psychologists and economists have studied a related phenomenon, projection bias, for some time: our tendency to assume that future tastes will resemble current ones more than they really will. Some of the evidence is charmingly mundane. People buy more cold-weather clothing when the weather happens to be cold when they order it, and are subsequently more likely to return it. See also portable aircon units this summer.

The same effect appears in much bigger decisions. Research covering tens of millions of car transactions and millions of house purchases found that the weather at the time of purchase affected choices including convertibles, four-wheel drives and, in a rather obvious reveal for where this study took place, how much buyers appeared prepared to pay for houses with swimming pools or air con. It’s self evident that a hot day can apparently make a swimming pool look like a more enduring feature of the life you want than it will seem in February.

All of which is mildly amusing when such a mistake is an unused coat. It becomes rather more interesting when the thing lasts a decade or more.

Consider buying a new kitchen. You are choosing it for a household that will change while the kitchen remains literally and figuratively unmoved. Children get older. People work differently. Parents become less mobile. The person who currently enjoys time consuming or otherwise elaborate cooking may not always do so. The huge island designed around a convivial bunch of people eating together may eventually accommodate two laptops a week’s post and an abandoned bowl of bananas.

Yet most kitchen-buying journeys quite reasonably concentrate on establishing what you want now and in the near term. How do you cook? What do you like? Which finish? Which appliances? How much storage? Sensible questions all, but they nevertheless assume that the customer answering them is a reasonable proxy for the customer who will still be living with those answers ten years down the line.

It’s a problem we wrestled with on AGA as an even clearer example. It is an unusually emotional purchase with an unusually long physical life. Someone can be buying a particular idea of home as much as an appliance, while making assumptions about how they will cook, entertain, use energy and organise their household many years into the future.

Cars too compress the same problem albeit into a shorter ownership cycle. A customer configuring a £70,000 car measurable can spend a remarkable amount of time deciding on wheels, trim, seats and tech while making shakier assumptions about the things that will more reliably determine whether the car actually suits them: mileage, commuting, children, dogs, mobility, where they live or whether they can charge at home. In which circumstances a configurator is very good at helping them specify the car they desire today. It is less interested in whether the car makes sense for the life they are likely to have whilst owning it.

Perhaps then the broader problem is that businesses are generally better at designing for differences between customers than for changes within the same customer. We segment people, research their needs and optimise journeys around what they tell us. Then, at the point of purchase, we implicitly freeze them in place.

Now, that lens becomes particularly interesting when the organisation, as most of course do, measures success at the point of conversion. In the examples above, the customer has ordered the kitchen, signed the finance agreement or placed the deposit, so the data point shows the journey worked. But that only tells us that the system was effective at producing a transaction. It says nothing about whether the decision aged well.

By way of a thought experiment then, imagine putting the following question halfway through a checkout:

“Under what reasonably foreseeable changes in your life would this stop being a good decision?”

Reader, your optimisation or conversion teams may reasonably object and, asked literally, it sounds like the business has decided suicidally to introduce doubt at precisely the moment it should be trying to remove it. But it is not a literal suggestion but rather a usefully provocative question for the people designing the whole purchase system.

What future changes could make this purchase wrong? Which can reasonably be anticipated? Where should the customer be helped to think about them, and where would that simply create needless anxiety? Can the product or service adapt later if those circumstances change?

A kitchen company might instead therefore help someone consider how a space works as children grow up or mobility changes. A retailer of high-end cast iron stoves might talk honestly about seasonal use, running costs and how people actually live with the product over many years. A car journey could surface likely changes in annual mileage, charging access or family use before somebody has spent a Sunday morning deciding between two varieties of tan leather.

This is less a matter of adding another clever question to the website than deciding what a good outcome actually is. Sometimes removing uncertainty will help the customer. Sometimes the same uncertainty is telling you something useful about the decision. For an expensive, long-lived purchase, someone who has never seriously considered whether the thing will continue to suit them may be very confident at checkout and considerably less confident six months later. Whereas someone whose reasonable doubts have been explored and answered may take slightly longer to decide, but they may also be less likely to cancel, regret the purchase or feel they were pushed into it.

That may be commercially useful in ways a purist conversion dashboard struggles to see. The sale still matters, obviously. But how can we ensure also continues to look like a good one once the customer has become somebody slightly different?

AI: I used AI to do some research to freshen up my knowledge of the specific behavioural biases at work and to stop me circling the point too many times. It produced an extract of the post and the tag list. The words however are therefore mostly mine.

Tagged , , , , , , , , , ,
Sep 11 2026
work

Before Someone Signs Off on the Robots

Have you seen the new robotic parking at Gatwick? It looks quite cool.

It works like this: you drive up into a garage bay, get out, scan a code, take your luggage and wander off to get on a bus to the terminal. Meanwhile, as you bump along the perimeter roads on your way to Pret a Manger and a surly greeting at security, your car is being lifted by a robot and shuffled into a parking space “fully autonomously” and presumably without the clattering of doors into each other or a young lad with a vape doing power slides in a field.

It reminded me of another experience in innovation I recently revisited: Pod Parking at Heathrow.

Ten years or so ago I tried it on a business trip to Scotland. A pre-dawn drop-off in a largely empty car park (back then it was still newish and posh enough to feel business-only), followed by the novelty of a solo ride in a pod to the terminal.

Returning this year with my son on a ski trip, it was slightly less fun.

A busy car park hunting for a space. A long walk with bags to the station. Then the ignominy of sharing a pod with other people similarly bedecked with ski equipment and making inane small talk before setting off on a circuitous, stop-start loop towards the airport.

Notwithstanding the fact that I’m now in my late 40s, travelling with rather more family paraphernalia and rather less wide-eyed consultant enthusiasm for clever infrastructure, the whole thing instead felt like a technically neat attempt to solve the wrong problem.

But before I explain why, let’s return to Gatwick’s robots.

Apparently they’re constrained a little. A fact mentioned almost in passing in the voiceover of what I assume was a press video: there are limits on vehicle weight, wheel size and wheelbase.

Plenty of people wouldn’t think twice about that. I did.

The service is marketed as a premium solution and has a five-day minimum stay. So it seems reasonable to assume a fair proportion of customers will arrive in large, expensive family cars.

Let’s say you own an i7, a Volvo EX90 or perhaps you’ve ordered the Range Rover EV. None of those can use it.

Already this premium, cutting-edge parking solution can’t accommodate a fairly representative selection of premium, cutting-edge family cars. Edge cases, perhaps, but also edge cases sitting at the leading edge of a market that is getting heavier, longer and increasingly fond of enormous wheels. The 21-inch wheel restriction feels particularly egregious.

Of course, the robots can presumably be enhanced. But significant capital has already been sunk into the system. Does the machinery now keep pace with the cars, or do customers eventually discover that their apparently ordinary SUV has fallen outside the parameters of the futuristic car park?

In any event, I started wondering whether robotic parking is actually solving a consumer problem. Or, more accurately, whether it is solving the right consumer problem.

From Gatwick’s point of view it may make perfect sense. Cars can be packed closer together because people don’t need to open the doors. You need less circulation space. There may be fewer staff movements. Less risk of somebody clipping a wing mirror or rearranging the side of a Cayenne against a concrete pillar. Potentially you get substantially more parking capacity out of an extremely valuable piece of land. Those benefits are undeniable.

But you’ll note they are largely benefits to the airport and parking operator. The customer has a rather different problem.

They want to arrive, find where they need to go without circling three times, get kids and bags out, leave the car somewhere they trust is secure, and ultimately reach departures with the minimum possible faff.

Preferably, too, without standing in a bus shelter in horizontal sleet debating whether the matrix display’s ‘6 mins’ has in fact changed at all in ten minutes.

Seen that way, the robot is doing something very impressive several stages after the customer has stopped caring. Once I’ve walked away from the car, I don’t really mind how it gets parked. A robot can carry it above its head while bleeping like a jolly droid. My attention has moved on to the next thing: where is the bus, how long will it take, will there be space for the bags, and how long do I need to stand next to the pseudo-Love Island contestant in coordinated loungewear before I’m actually inside the terminal?

Which brings me back to Heathrow’s pods. The engineering proposition is, or at least was, immediately seductive. Small autonomous vehicles, ostensibly on demand, with a dedicated congestion-free route to the terminal. The PowerPoint similarly likely wrote itself. Judged narrowly as a transit system, it’s evidently clever.

But the judgement should be from the passenger’s experience. Here the journey begins not when the pod moves but when you stop your car: park. Extract luggage. Work out the nearest station. Walk in the rain. Wait. Board. Travel. Sometimes stop — “We will resume your journey shortly”. Get out. Walk again with luggage. Find the terminal entrance.

That is the system and, when closely observed, some fairly boring alternatives instead appear surprisingly competitive.

Put more bus stops alongside parking rows. Increase their visibility and directional signage. Cover the walking routes (maybe even with solar panels). Tell people exactly when the next bus will arrive. Run enough buses that nobody really needs to think about the timetable at all.

And make the bus itself good. Warm in February, properly air-conditioned in August. Comfortable seats. Proper luggage racks designed around what people actually take to airports rather than a grudging shelf above the wheel arch. A driver who says hello, helps someone with a heavy case and knows where they’re going. Give the thing a bit of design attention.

None of that is technologically remarkable, but I suspect the combined effect on the journey could be disproportionately positive.

You’re unlikely to find an architectural visualisation of a man standing happily beneath a bus shelter on the homepage of Dezeen. Or a breathless article about wayfinding signage in Monocle.

But it may produce a better experience.

Perhaps, too, there is another behavioural wrinkle. Five minutes of waiting for something you cannot quite predict can feel substantially worse than eight minutes of movement towards your destination. Walking, as many have observed, gives you a sense of progress and control.

Although luggage complicates the calculation.

A 400-metre walk without luggage can feel like useful progress. The same 400 metres dragging two cases, a ski bag and a slightly recalcitrant child across coarse tarmac is not quite the same behavioural proposition. At some point movement ceases to feel like autonomy and simply becomes work.

Which is why shaving two minutes from the vehicle journey may be almost worthless if the passenger has had to drag their luggage an extra 400 metres to reach it. A beautifully optimised autonomous transit system can quite easily lose to a diesel bus simply because the diesel bus stops adjacent to your car.

So often, sophisticated experience design goes slightly wrong along such lines. Organisations naturally optimise the things they can see, measure and control. In this case: vehicles per hectare, average transfer time, labour requirement, throughput, number of drivers, perhaps even CO₂. All sensible metrics.

People, rather inconveniently, experience the gaps between them. They experience the anxiety of where to go. The extra walk nobody included in the journey-time calculation. The toddler who needs a wee. The suitcase with a wobbly wheel. The bus that pulls away just as you approach the stop. Such things are difficult to measure, don’t align neatly to one system or department and so disappear from optimisation models altogether.

So when we end up with technically excellent things that somehow feel worse, I wonder if there is also an element of theatre at play.

Airports have always been preternaturally good environments for conspicuous modernity. The driverless pods, biometric gates, robots and baggage systems, vast digital screens. They are signals, often at a national identity level, of investment, progress and future-facing willy-waving.

A covered walkway, even with solar panels, is not that. Neither is a bus stop being moved 80 metres. Yet either could improve more journeys, more often, for less money.

None of which is to denigrate the robots too much. They’re far from pointless and if Gatwick can materially increase parking capacity without tearing up more green fields, robotic parking may prove to be quite the investment. Judge it on density, operating cost, reliability and land use and the argument may be very strong. But then call it what it is: clever parking infrastructure.

The more useful question is simply whose problem we’re actually solving. Sometimes installing a fleet of robots that lets you cram more cars onto the same patch of land genuinely does improve the customer experience. Sometimes the customer would get rather more from a bus stop 80 metres nearer their car, an accurate countdown and a warm bus with somewhere sensible to put the cases. The difficult part of experience strategy is having the confidence to work out which before someone signs off on the robots.

AI: I used ChatGPT to tidy up some grammar and that, dear reader, is all.

Tagged , , , , , , , , , , , , , , , , , , , , , , , ,
Jul 07 2026
socio-cultural, work

Belgravia, Overheard

I’ve had a busy few months which, from a financial perspective at least, has been no bad thing. The downside is that the writing has rather suffered in frequency. Oddly though, I find the opposite happens in my head. When work is busy I seem to notice more. Every train journey, client meeting, conversation and aimless wander through London gets mentally filed under “there’s probably a piece in that.” The vast majority stay there.

During quieter periods, visibility anxiety has an unfortunate habit of convincing me that almost every observation deserves to become a LinkedIn post. I don’t think I publish complete nonsense, but there is probably a regression to the mean. Which is all to say that this was one of the ideas that escaped the notebook. You can decide whether it should have.

I’ve always liked the contrast between elite athletes and their coaches. You see someone with the physique of a middle-aged middle manager calmly explaining to a twenty-two-year-old Olympic hopeful why their hips are opening half a degree too early. It feels faintly absurd but of course the coach isn’t there to run the race. They’re there because they’ve developed the judgement over time and perhaps their own career to notice things the athlete no longer can.

I occasionally wonder whether writing about luxury is a bit like that. Unlike some of LinkedIn’s more glamorous commentators, I’m not bouncing between Aman resorts, Formula One paddocks and yacht launches in Como. My natural habitat remains Esher, the quiet south Cornish coast, RHS Chelsea and occasionally Regent Street if I’m feeling particularly in need of retail.

Then again, most athletics coaches aren’t still running the 110m hurdles.

A few weeks ago I found myself walking through Belgravia. It’s one of the few places in London where almost every version of wealth seems to coexist. You see extraordinary money, of course, but what I noticed wasn’t the value of it but it was that everyone appeared to be speaking slightly different ‘languages’.

A Ferrari slipped past me on Audley Street. I didn’t know what it was, as is typical it carried no model badge. On Burlington and Bond Street there are shops where the bags in the window carry no logo or at least logos so small as to be unreadable from a distance.

And then, a few streets later, came a Rolls-Royce Cullinan in a vulgar magenta with the kind of chrome accents that wouldn’t be out of place on a 50s Cadillac.

None of the people owning these goods, I suspect, would consider themselves to have anything other than luxury tastes and certainly not bad taste. They would consider themselves discerning. Yet each are signalling entirely different things.

To return to Ferrari. I’ve always found it fascinating that a company capable of building a bright red, mid-engined supercar, loud enough to be described as “an operatic rhinoceros trapped inside a tumble dryer”, draws the line at six scripted letters across the boot line. Reader, it is many things, but subtle isn’t one of them.

And yet somebody inside Ferrari decided that “296 GTB” didn’t need telegraphing. If you know, you know. If you don’t, a badge ain’t going to change you.

Porsche on the other hand seems to have reached the opposite conclusion. Their model names have that daft run-on suffixing , but worse than that is their love of decals and special edition badging and stitching. Range Rover briefly lost its mind too. During my time around the brand it seemed every derivative acquired another flourish of nomenclature until the back of the car resembled one of those late-90s PCs covered in stickers announcing Intel Inside and “Designed for Microsoft Office.” BMW, meanwhile, has scattered M badges across the range with the enthusiasm of a toddler decorating a birthday card.

None of which makes them wrong, but it does make me wonder whether luxury itself is beginning to diverge. There is one version wants to be recognised and another that only needs to be recognised by the right people.

At this point somebody with a proper strategy background and much cleverer than me would probably invoke Pierre Bourdieu, so I’ll save them the trouble. Bourdieu argued that status isn’t simply about money, but rather what he called cultural capital: the accumulation of taste, knowledge and judgement that subtly tells other people where you belong.

Owning an expensive watch tells me you have money. Recognising an expensive watch that almost nobody else notices tells me something rather different. One is a demonstration of wealth, the other a demonstration of fluency. Unfortunately, Instagram rather complicated the whole affair.

Online, luxury became something performed for people scrolling past in fractions of a second. In this environment logos and monograms work brilliantly. A Louis Vuitton bought in Tenerife because everyone back home will recognise it. The Balenciaga hoodie bought on Vinted because the logo is more important than the drop-season, also works.

At the opposite end of the wealth spectrum, exactly the same instinct produces superyachts like shopping centres and watches that look capable of summoning a medevac. Different bank balances, remarkably similar behaviour.

The trouble is, cameras are hopeless at photographing ambiguity and judgement; design lines and proportions rarely survive a thumbnail. Neither does impeccable tailoring nor hospitality that is anticipatory and personal rather than performative. My vision of luxury might be a beautifully balanced Hoek yacht easing through Scandinavian waters while the owner wears a twenty-year-old jumper because they genuinely couldn’t care less. I covet the brands and experiences that show curation and restraint.

And before anyone starts sharpening the comments, I don’t think one version is morally superior to the other. Vulgarity, after all, is culturally relative. To the Emirati collector, a two-tone Phantom with intricate marquetry might represent extraordinary craftsmanship and generosity. To a Nordic architect it may feel grotesque. Neither is objectively right, they they’re simply speaking different aesthetic languages.

Diagnostically, I wonder whether we simply misunderstand the people buying these things. Which speaks to my own scepticism and misgivings of traditional personas in luxury.

Male.
Fifty-four.
£250m family office.
Owns homes in London and Verbier.

Fine.

Now tell me whether he commissions a floating palace with a cinema, helipad and c., or spends the next decade restoring a Swan 65 because the joinery is beautiful and he likes crossing the North Sea under sail. Those are completely different people in a way that doesn’t show up demographically but culturally.

Traditional segmentation along these lines has remarkably little to say about such distinctions because it’s looking in entirely the wrong place. So, as I jumped on the train home that warm evening in London, I jotted down a few thoughts that tried to unpick why I still see brands spending a great deal of time asking who customers are: their age, income, their geography and profession, perhaps so anchored in media buying. Whereas I believe in this sector the more insightful question is something else entirely. Who do they imagine they’re speaking to?

Because somewhere between that clean-lined Ferrari and the bling-trimmed Cullinan, it occurred to me that luxury isn’t one language or culture at all. It’s a collection of dialects.

Tagged , , , , , , , , ,
Apr 27 2026
work

The outsider

The room is already settled by the time they speak. Those slides have been through two internal reviews. The language is clean, defensible, already halfway to sign-off. You can feel the work closing in on itself, decisions calcifying because they’ve been seen often enough to feel inevitable not necessarily because they’re right.

Then someone arrives who hasn’t been part of that loop. Sometimes it was the agency founder, drifting in late, coat still on, just finishing a call, carrying that slightly impatient calm of someone who’s seen the pattern before. Sometimes it was the strategy partner who’d been impossible to book, sitting down quietly, eyes half closed, the fug of coffee doing its work, before cutting through the thing in a sentence that made the rest of us wince. Occasionally it was the ECD, flicking through printouts, pausing just long enough to drop a few pages on the table. No speech. Just a clear, deliberate gravitational rejection of what we’d all convinced ourselves up to that point was good.

But perhaps just as often, it wasn’t experience doing the work. It was the person who shouldn’t really have been there. The graduate. The outsider. The one invited for texture and politeness rather than judgement. The one who hadn’t learned the language yet, and so couldn’t follow the careful logic the rest of the room had built. They’d ask something slightly misplaced. Pull in an example that didn’t quite belong. Miss the point, on paper. And in doing so, expose it.

These effects are rarely dramatic, there’s no grand pivot. No theatrical breakthrough. Just a tilt. A detail that no longer seems to hold water. A narrative that suddenly feels a bit too clever. Enough to loosen the grip the idea had on the people in the room.

Good teams make space for that kind of disruption. Without being theatrical or insisting on “fresh perspective”, but rather as a way of testing whether the work can survive contact with someone who hasn’t been trained to agree with it.

Most ideas don’t fail in the room. They fail later, when they meet someone, often client-side, who doesn’t share the context, the patience to go through all the supporting slides, or the goodwill.

The value of the outsider isn’t that they’re right. It’s that they haven’t yet learned how to be politely wrong in the same way as everyone else.

In the past six months, I’ve found myself brought into rooms both virtual and physical for exactly that reason. Not to lead the work, or to own it, but to sit slightly outside it. Close enough to understand the logic. Far enough away not to be bound by it. Sometimes that has meant drawing on experience. Sometimes it’s just been asking the question that feels a bit off, or pointing at the thing everyone has quietly stepped around.

The aim isn’t to derail the work. Just to tip it, slightly. Long enough to see what holds its balance, and what doesn’t.


Tagged , , , , , , , , , , , ,
Apr 01 2026
work

I went back to Uni. What I learned on my Human-AI Interaction course.

The bit that mattered was never the screen.

Take this example. You’re looking to buy a used car. You shortlist three cars and start nudging the finance up and down. Nothing dramatic, just enough to see what an additional £30 a month buys you. A slightly newer plate perhaps. Fewer miles. A bit more punch, or practicality.

Then you click through to a retailer. Your shortlist disappears. Your modelled finance examples disappear. You’re on a blank form, being asked questions your behaviour has already answered. Nothing failed in any obvious way. The system simply dropped the ball precisely at the point it needed to hang on to it.

That is still how a great deal of digital experience works.

You can now generate a respectable interface in seconds, in Figma Make or hell even Gemini. It’ll be clean, structured, and broadly acceptable. Enough hierarchy to feel coherent, enough restraint to avoid embarrassment. It looks like design, which is why people are getting carried away and, frankly, why people are being laid off.

There is too a growing belief that interfaces are becoming fluid. Instead of designing fixed screens, we define components and let agents assemble the right interface on demand. Ask for something, get a form. No navigation, no structure, just translate intent directly into interaction.

Concepts like this are gaining column inches, conference talks and LinkedIn views because they’re brief and intelligible. However, as soon as the interaction extends beyond that moment, the weaknesses become apparent. Context isn’t carried forward properly. Earlier decisions aren’t respected. The system produces something plausible but slightly off, and the user ends up repairing it, effectively doing the work we’d hoped to have eliminated. The surface has improved. The underlying behaviour hasn’t.

Over the past eight weeks, I’ve been working through a Human–Computer Interaction and AI course with the University of Cambridge (Advance Online), and it has been mildly uncomfortable in the way good things often are. It stripped things back to a version of UX that feels almost unfashionable: define the problem before you touch a solution; model the system in terms of what it actually does, not how it looks; decide where control sits between human and machine and how that moves about; draw a boundary and accept responsibility for what happens inside it; then prove the thing works rather than just assuming it does.

As I worked through the modules, it was charmingly familiar and obvious that none of this was new. It just hasn’t been particularly visible for a while. The industry has been busy polishing UI surfaces and calling it progress. AI hasn’t changed that per se, but has made the gap harder to ignore. You can now generate something that looks finished without having done any of the thinking that would make it hold together.

That’s also why a lot of current AI design work feels slightly misplaced, and why the obsession with UI product designers irritates me. There are people doing very good work on interfaces and component systems. Zander Whitehurst is one of them, and they look excellent. But that work sits downstream of where the real difficulty now is. You can refine the surface as much as you like; if the system doesn’t carry intent, it won’t survive contact with actual use.

The “democratisation of design” line doesn’t stand up to the faintest scrutiny. Sure, more people can produce interfaces, and there’s been a lifting of all boats in terms of aesthetics. That’s true. Almost none of those interfaces are grounded in any understanding of what the user is actually trying to do. They meet a baseline. They look competent. They solve very little.

What has been democratised is production.

The discipline that underpins it, the work of defining what matters, what to surface, what to hide, what to carry forward, and where the system should stop and ask, has not been automated. It has been skipped.

The shift to intent-based interaction makes that gap more obvious. You are no longer stepping through a process one action at a time. You state an outcome, and the system attempts to get you there. That changes the shape of the problem. The system has to interpret intent, apply constraints, decide what to do next, and show enough of its reasoning that you can tell whether it has misunderstood you. When it gets that wrong, it doesn’t look like a broken interface. It looks like a reasonable answer that doesn’t quite fit, which is both harder to spot and harder to recover from.

This is where the work is moving.

If interfaces can be assembled on demand, the value shifts into what sits behind them: how intent is captured, how memory is handled, what the system is allowed to assume, when it must ask, and how it behaves when it reaches the edge of its understanding.

Friction becomes part of that. For routine actions, speed is fine. For anything with consequence, removing every pause produces something that feels smooth and behaves carelessly. A system that never slows you down also never asks you to think.

There is a second effect that is easier to miss. If the system does more of the execution, the user does less of the thinking that used to go with it. Over time, that changes behaviour. Trade-offs become less visible. Assumptions go unchallenged. Outputs are accepted because they look plausible..

You end up with users who are comfortable approving things they don’t fully understand.

The risk here isn’t replacement. It’s avoidance. If the problem isn’t defined, if the constraints aren’t understood, if the system isn’t designed to carry intent across time, then the speed of output doesn’t help. It just produces more surface, faster, over the same unresolved issues.

That is already visible. The used car journey doesn’t fail because the interface is ugly. It fails because no one took responsibility for the whole. The same pattern is now being reproduced with better tools and more convincing output.

If that continues, the work doesn’t disappear.

It just gets left to the user.

AI: I used ChatGPT to tidy up some grammar and I used Gemini to pick some holes in the piece to strengthen the arguments.

Tagged , , , , , , , , , , , , , , , , , , ,
Mar 04 2026
socio-cultural, work

Is 40% of Work Unnecessary?

Back when I was a full-time employee I can distinctly remember, like many of us I assume, looking at my calendar and thinking: this cannot all be essential.

The client work was demanding, but it was real. Framing problems. Untangling clumsy journeys. Making regulated systems usable. That part didn’t trouble me. It was what I always wanted to believe I was actually being paid to do.

The Sunday night dread was never about the difficulty of the UX tasks I had to do; it was about the prospect of defending a button placement to three people whose job titles I didn’t fully understand.

That feeling was neither rare nor dramatic. It was simply ambient.

Minimal home office desk in soft winter light, with a laptop, an open weekly diary filled with handwritten notes, and worn leather gardening gloves beside a window overlooking a blurred garden hedge.

When a YouGov poll reports that 37% of UK workers felt their job didn’t make a meaningful contribution, and a similar Dutch poll finds it at 40%, it therefore feels somewhat familiar.

David Graeber called them “bullshit jobs.” His argument was broad. Mine is narrower. Most knowledge roles contain a necessary core wrapped in a layer of activity that grows over time. Reviews of reviews. Meetings about alignment. Artefacts written, designed and built to signal diligence in output. Essentially, as Clarkson used to say “I did a thing!“.

None of it looks absurd taken in isolation. Together, it grows.

For more than twenty years I worked five days a week in agency life. The rhythm was typically conventional: full calendars, visible participation, institutional choreography. It signalled seriousness. It was still frequently fun around the margins. It also created drag.

In late 2024 I left, and as I found my new life as a freelancer I was able to cut the week to <5 paid days.

I didn’t change the nature of the work but I have changed the size of the container.

The client conversations still happen. The user experience architecture still happens. The thinking still happens. What has (mostly) disappeared are the rituals that demonstrated I was thinking. Those pre-reads, two hour workshops on alignment. Status updates with no status updates. The sorts of meetings that PMs get all excited about or shitty with you when they can’t move their Trello/GANT along a notch.

The surprising part was how little of that missing activity altered the outcome.

On the other days I study human–AI interaction (more on this in the weeks and months to come) and I work at our allotment. Recently I completed the task of laying a hedge in a centuries-old craft. Laying hedge is slow and slightly brutal on the hands. You cut partway through the stem, bend it, weave it into the next. At the end of the afternoon there’s a line you can lean against. If it fails, you can see the weak point, and if you can’t, a sheep probably can.

There is no meeting about the hedge. It either stands or it doesn’t.

That contrast isn’t noodling pastoral escapism. It simply highlights something about knowledge work: when outcomes are abstract and time is fixed, activity expands to reassure everyone involved. AI now presses on that reassurance.

A large share of white-collar work involves drafting, structuring, summarising and formatting. My wife and brother are both lawyers and I can see it in the sheer volume of their workload. These tasks will become faster with machine learning and ai modelling. One person can carry more of them. That has and will create slack.

Organisations can therefore reduce the same container I did. Or they can preserve the container and invent new layers to supervise the tools that removed the previous layers.

The exact percentage (whether 40% or otherwise) is less important than the recurring personal question: “If we stopped doing this, what would actually fall apart?” Most people can answer that for at least part of their own week.

The difficulty isn’t identifying what’s surplus work. It’s that that surplus seems to underwrite role hierarchy, salary bands, and the whole shared fiction of the full week. Remove it and you have to redesign more than diaries.

Cutting my own week didn’t change the work I did. It changed what was visible. Once you notice how much of a calendar exists to justify itself, it’s difficult to stop noticing.

AI: Just a bit of Grammarly used to tidy up the copy and some ChatGPT to surface some references and generate a (rubbish) image because algorithms.

Tagged , , , , , , , , , , , , , , , , , , ,
← Older posts