Showing Them the Future
How a working prototype turned "can this even be done?" into "when can we have it?"
The meeting room was full of men who had spent more years in insurance than I had spent being alive. It was April, and the higher-ups wanted the whole thing decided in April. On the table sat my laptop, and on the laptop sat a prototype, a clickable ghost of a product that existed nowhere else. Behind it there was nothing. No code, no database, no development team assigned. I was the most junior person in the building, by a margin nobody bothered to measure, and I was there to show a table of veterans whether something they quietly doubted could be done at all.
I had joined a software services company in 2022 as a junior designer, mostly because my portfolio was thin and I wanted one serious project in it. Within a few months I was carrying four at once, which is normal enough for a junior at an agency. Then the senior designer fell severely ill and had to travel abroad for treatment, and there was no second senior to absorb what he left behind. Three of my projects I could keep on their feet. The fourth was the one I quietly dreaded, the hardest of them by a long way, and it is the reason I am writing any of this down.
The client was a prominent, long-established insurance company, the kind of institution where decisions move slowly and paper moves slower. They wanted to modernise their website, and the centrepiece of that was a set of premium calculators. A premium calculator is a small, unglamorous tool. A customer enters a handful of facts about what they want to insure, and the tool returns what the policy would cost. Described like that, it sounds like an afternoon of work. In reality it was several completely different products wearing the same plain face, each with its own logic and its own exceptions, and all of that logic lived in one place, the company’s old paper documentation of its insurance rules, pages photocopied so many times the letters had gone soft at the edges. I doubt anyone in that building had read the whole thing in years. To design something that actually held together I would have to, so I spent the next several weeks doing exactly that, reading every page.
That reading is where the project stopped looking simple. Take motor insurance, the deepest of the calculators. Before a single price can appear, the logic splits on whether the customer wants comprehensive cover or only act liability, which in plain words is the difference between protecting the car itself and carrying the bare legal minimum. Then it splits again on whether this is a first policy or a renewal, then on what kind of vehicle it is, a private car, a commercial vehicle, a motorcycle, because each category reads from a different part of the rulebook. From there it wants the vehicle’s price, the number of passengers, the dates the policy will run, whether the person behind the wheel is the owner or a paid driver, whether the vehicle carries a tachometer or a tracking meter. And then comes the strangest part, a menu of risks the customer can choose to give up. Decline cover for riot and strike damage, for flood and cyclone, for earthquake, and each refusal shaves its own specific fraction off the rate, and several of them can be stacked together. What a customer would meet as a short form was really a decision tree with dozens of branches, and each branch traced back to a line somewhere in those soft photocopied pages.
The other calculators were different animals with the same temperament. Cash-in-safe insurance, the cover for money kept in a vault, cared who you were, a bank, a business, a private company, and where the building stood, and how much cash the vault could hold, and which category the location fell into for what the documents called Strike, Riot and Civil Commotion, the clause covering damage from unrest. Then it asked a question I have never forgotten, because it felt like the rulebook reaching out to touch the physical world. What is the building made of. Full concrete counted as one class, brick walls under a metal-sheet roof as another, full metal sheet as a third, and each pulled a different number out of the book, because a vault is only as safe as the walls around it.
This was the texture of the whole assignment, hundreds of small, stubborn rules written for clerks, now needing to be reshaped into something a stranger could walk through alone, with the machinery kept out of sight. And this, I slowly understood, was exactly what made the people above me hesitate. The officials I would eventually face were largely in their sixties. They had administered these rules across whole careers, and they knew better than anyone how tangled the thicket really was. Their hesitation was a form of expertise.
There is a name for what was happening in that hesitation. In 1961, Daniel Ellsberg described something odd and very human about how we handle the unknown, an effect now called ambiguity aversion. Offered a gamble with known odds and a gamble with unknowable odds, people reliably take the known one, even when the mystery might pay better. What truly unsettles us is the bet we cannot see, the one with odds nobody can name. The room I was preparing for was a live demonstration of it. Nobody could put odds on this project. The thing being asked to exist had never existed there before, the only person holding both the rulebook and the design in one head was a junior, and the cost of finding out was a real team for a real stretch of months. They had a harder question than whether they liked any particular design, and it was the one holding everything still. Could this much complexity become a usable tool at all.
Against that kind of fog, a description is worthless. You can stand in front of cautious people and talk about an unbuilt thing for an hour, and to a skeptic the words are just weather. So it had to be shown. I built a working, clickable prototype, end to end, every calculator, every branch I had pulled out of the paper. Then I sat in meetings with men who could have retired before I learned to write, and I walked them through it. Pick a vehicle. Choose comprehensive. Decline the flood cover and watch the rate move. Nobody had to take my word for anything. There was a thing on the table, and it either held together under their hands or it did not.
It held. And somewhere in those meetings the question changed. It stopped being whether this could work and became when they could have it, which is the entire distance between a stalled project and a funded one, crossed without a single line of production code being written. That is when I understood what I had actually been making all those weeks. The prototype’s real product was never the screens. It was certainty. The ideas were there, and engineers could be on it within a week. What no one had been able to buy, at any price, was confidence, and a prototype is a machine for producing exactly that. It turns an unknowable risk into a visible one, something you can poke at and put a number on, and that is what let’s careful people say yes.
I think about that spring often now, because the part of the work that produced the screens is exactly the part that has since become cheap. Code can be generated. Layouts can be generated. If the building had been the valuable part of what I did, that value would already be gone. But the building was the last and least of it. What could not have been handed to anyone, or anything, was the months of understanding underneath. Reading a rulebook nobody else would finish, holding all of its branches in one head, knowing which complexity the customer needed to see and which needed to disappear, and knowing the room well enough to turn all of that into a calm walk-through for people with every reason to doubt. The scarce work, then and now, is the understanding that turns a risky fog into something people will commit to. I say this with no anxiety about machines. It is simply where the difficulty lived.
The same machine has a darker setting. A prototype can manufacture confidence it has not earned. A slick demo for something that cannot actually be built is its own kind of lie, and a persuasive junior with a laptop is, in the wrong hands, a dangerous instrument. I was improvising at the edges, and I knew it. What made the confidence in that room legitimate was that every screen they clicked traced back to a clause I had actually read, and every branch of the logic was real before it was pretty. The certainty I sold them was certainty I had gone and gotten myself, first. Let those two things drift apart, and the same tool that earns trust quietly starts to steal it.
Once you see this pattern you find it everywhere people try to move cautious institutions. The founder pitching a company that is mostly conviction and slides. The architect who builds a scale model so a committee can lean over a future block of the city. Each of them is performing the same conversion I stumbled into that spring, turning somebody else’s ambiguity into a risk that can be looked at, and the future tends to go to whoever does that honestly, more often than to whoever merely had the best idea. Making an unbuilt thing feel real and safe enough to commit to is its own discipline, and I have come to think it is a steadier one than most of what we call skill.
The strangeness of that April has never quite worn off. The youngest person in the building, holding a thing that was not real, lending a room of veterans the confidence to make it real. The calculators got built, and like all software they have been changed since and will eventually be replaced. The understanding that let everyone begin is the part I still carry, and I suspect it was the part that mattered all along.