My first job involved making brochures you could turn the pages of. Twenty years on I still think they were a bit of a gimmick, and I have started wondering whether we are doing the same thing again.
The Page That Turned
My first proper role was building ebrochures at a production agency. If you have not come across one, the idea is simple enough: take the printed brochure a retailer puts in their stores, and put it on a screen so you can flick through it.
Not scroll. Flick. You clicked the corner of the page and it peeled over, with a shadow that followed the curl and a little papery noise as it landed. Some of them had a sound on the spine when you opened it.
I thought it was brilliant. I was in my twenties, I had just helped put a shop's worth of product onto a computer screen, and it did something nobody in the office had seen before. Clients loved it. You could show it in a meeting and watch people lean in.
I am still slightly fond of it, which is worth saying before I take it apart.
Did It Sell Anything? Probably Not
Here is the bit I have had two decades to think about.
I do not believe those ebrochures sold more product. I have never seen anything that convinced me they did. What they did was let somebody look at the brochure without going to the shop, and that was a real thing worth having. Plenty of people could not get to the retail space, or could not be bothered, and now they could see the range anyway.
So there was value in it. Just not the value we were selling. We were selling the page turn, and the page turn was decoration.
The bit that has aged worst is the reading experience. A printed spread is designed to be held at arm's length, two pages at a time, at a size somebody chose deliberately. Put that on a phone and you get an enormous amount of wasted space, text too small to read, and a click-to-turn interaction on a device whose entire design language is the scroll.
The content was never adapted to the channel. It was transplanted into it.
So Why Are They Still Here?
Because they still are. You will find them today, on real sites, from real companies. And I do not think that is because anybody sat down and concluded they were the optimal way to present a product range.
I think it is two things.
They are effectively free. The printed brochure is the thing the business actually makes. It gets designed, signed off and paid for anyway. Turning that finished artwork into a flipbook costs approximately nothing, and nobody has to justify a budget for it. It is a by-product.
They are familiar. Everybody knows what a brochure is. You do not have to teach anyone how it works. There is no learning curve, no explanation, no risk that a customer opens it and does not understand what they are looking at.
Free and familiar. Notice that neither of those is "it works better". They are reasons to do something, and they are perfectly reasonable reasons, but they are not evidence.
New Thing, Old Costume
The reason I have been chewing on this is that it is not really about brochures.
When something genuinely new arrives, our first instinct is to dress it in the clothes of the thing it replaces. Early cars looked like carriages with the horse removed. Software still saves to a picture of a floppy disk, an object most people using it have never touched. My phone's notes app used to have a little torn paper edge at the top.
There is a proper word for this and the design world has been arguing about it for years, but the useful version is simpler. Wrapping the new thing in the old thing's shape buys you adoption. People recognise it, so they use it. That is a genuine benefit and I am not sniffy about it.
The cost is that you inherit the old thing's limitations along with its familiarity, and you give up whatever the new medium could actually do. The flipbook cannot be searched properly. It is miserable with a screen reader. It cannot reflow to the device, cannot link out sensibly, cannot tell you what people looked at. All of that was available. It was traded away for a page turn.
The costume is meant to be a bridge. The problem is that bridges are supposed to be temporary, and this one has been up since about 2005.
Which Brings Me to the Face
I have been building things with AI for the past year or so, for clients and for myself, so what follows is aimed at me as much as anyone.
Look at how AI is being presented at the moment. It has a name. Increasingly it has a face, or an avatar, or a photorealistic video person who blinks and tilts their head. It has a personality, a tone, sometimes a backstory. Voices are being given hesitations and breaths on purpose, so they sound less like a machine. There are products sold as AI employees, with job titles.
We are putting the new thing in a human costume. And I keep wondering whether the face is this generation's page turn.
The parallel holds uncomfortably well. It is cheap: we already have twenty years of interface conventions for talking to a person, so pointing the technology at a chat window costs nothing to design. And it is familiar: everybody already knows how to have a conversation, so there is no learning curve.
Free and familiar. Same two reasons. Still not evidence.
What the Costume Costs
This is the part I actually care about, because with the flipbook the cost was some wasted space and an awkward click. Here it is a bit more than that.
A human shape invites human trust. We are wired to extend a certain kind of credence to something with a face and a name that speaks confidently in the first person. That is not a bug in the user, it is the whole reason the design works. But it means the interface is actively encouraging you to over-trust a system that is confident and not reliable, which is the exact failure I wrote about in the human-in-the-loop piece. Only there it crept in through familiarity. Here we are building it in deliberately.
A face implies things that are not true. It suggests memory, understanding, and someone who is accountable if it goes wrong. People calibrate to the costume rather than to the thing wearing it, and then get caught out when it turns out there was nobody in there.
It hides the useful bits. If I am going to rely on something, what I want to know is where this came from, how sure it is, what it looked at, and what it could not find. That information is exactly what a well-designed instrument shows you at a glance. It is exactly what a confident paragraph of prose from a friendly avatar buries.
It wastes the medium. Same as the flipbook. A conversation is narrow and linear: one thing at a time, in order, at the speed of reading. Software can show you six things at once, let you compare them, let you go back and change an input and watch everything downstream move. We have got a technology capable of that and we are mostly serving it through a text box with a name on it.
So What Should It Look Like?
I do not have a tidy answer, and I would not trust me if I gave you one. But there are directions that feel more honest than a face.
Intelligence that sits inside the tool you are already using, rather than a separate thing you go and talk to. Interfaces that show their working: what was consulted, how confident it is, what it is unsure about. Something closer to an instrument you read than a colleague you interrogate.
That is a bias of mine, admittedly. I have spent my career building operational systems, and I would always rather have a dial that tells me the truth than a voice that tells me it is fine.
But I notice that when I use these things properly, for real work, I end up wanting less personality rather than more. I do not need it to be charming. I need to know whether it is right.
Where I Would Push Back on Myself
A couple of things stop me being too pleased with this argument.
The ebrochure did do something useful, and I said so. Reaching people who could not get to the shop was worth having, even if the page turn was decoration. The familiar form was doing a real job.
The same is true here. A conversation is a genuinely accessible interface. If somebody cannot or will not learn a complicated piece of software, being able to just ask for what they want is a proper win, not a compromise. I have watched people who would never open a dashboard get useful things out of a chat box. That matters more than my preference for dials.
And the familiar form as a deliberate bridge is a legitimate strategy. One of the tools I built recently runs entirely inside a messaging app people already had on their phones, precisely because asking them to download and learn something new would have killed it stone dead. I made that choice for exactly the reasons I have just been sniffy about.
So the question is not whether the familiar form is allowed. It is whether you are using it as a bridge, or as a substitute for thinking about the medium at all.
The flipbook stopped being a bridge about fifteen years ago. Nobody noticed, because it was free.
The Question I Keep Coming Back To
I might be completely wrong about this. It is entirely possible that in ten years the human presentation turns out to have been right all along, and I will be the person who thought the page turn was silly while everyone else got on with it.
But I keep asking the same thing, about the brochures then and the avatars now: is this shape here because it is the best way to do the job, or because it was the cheapest thing to reach for and the easiest thing to explain?
With the brochures, I know the answer. I made them.
Building something with AI in it and not sure the interface is doing you any favours? That question sits right in the middle of our context engineering work. Give me a shout if you fancy talking it through.