At 2am on a Tuesday morning, somewhere between a half-formed idea and a cup of coffee, I opened my laptop with a simple question in mind: what if investment advice and investor education were not separate things?
Not a PDF sent after a meeting. Not a static proposal. But one integrated experience.
By 4am, I had a working prototype. Not a mock-up, not a set of slides, but a functioning application. Within a week, that prototype had evolved into a "Portfolio Prompt Builder" system capable of generating structured, advisor-grade investment proposals while simultaneously guiding users through a short, interactive learning journey explaining the underlying logic — from asset allocation to market timing and compounding.
That alone might sound impressive. But it is not the story.
The real story is how it happened.
There was no traditional development process. No requirements documentation, no backlog, no sprint cycles. Instead, what sat at the centre of the exercise was something much more fundamental: a structured body of domain logic. The same logic that typically sits embedded within private banking advisory processes — risk profiling, allocation rules, portfolio construction constraints — was translated directly into a functioning system using generative AI. Prompts became specifications. Specifications became execution.
A few days later, I ran a second experiment.
"This is the most mind-blowing thing I’ve seen in my 30 professional years," I found myself thinking — not lightly.
This time, I took the same structured advisory logic and fed it into a generative development environment. Thirty minutes later, I had built an "Investment Decision Lab". Not a mock-up, not a prototype, but a functioning application.
Again, there was no classical process. I didn’t write requirements or define user stories. Instead, I described the system in a structured prompt. The environment suggested practical enhancements in real time, and the system evolved as I was building it.
What emerged was not a toy, but a fully structured investment process: it validates inputs, generates investor profiles, constructs asset allocations — this time 100% rule-based and deterministic — provides rationale, highlights risks, simulates stress scenarios, analyses fees, and runs Monte Carlo projections.
The entire advisory logic was operational.
And yet, even this is not the story.
The story is that the bottleneck has moved.
For decades, software development has been constrained by implementation. Ideas needed to be translated into code, tested, integrated and deployed. Each step introduced friction, time and cost. As a result, product development cycles stretched over months, often years.
Now that constraint is beginning to shift.
The limiting factor is no longer primarily coding. It is the ability to structure domain logic clearly — to define rules, constraints and workflows in a way that can be operationalised. If you can express a system precisely enough, it can increasingly be built almost instantly.
In financial services, the implications are immediate. Investment advice has long been positioned as a human, relationship-driven capability. But much of what constitutes advisory — risk profiling, asset allocation, portfolio construction — follows structured logic. Once that logic is clearly defined, it can be codified, tested and reproduced.
At the same time, a second shift becomes visible. Education moves into the product itself. Rather than separating advice from understanding, systems can now guide users through the reasoning behind decisions as they are made. The product does not simply deliver an outcome; it explains itself.
This is where value begins to shift.
For years, the constraints were obvious: coding capacity, delivery pipelines, IT backlogs. These are no longer the primary limiting factors. The new constraint is structured thinking — the ability to define constraints clearly, encode decision logic and design coherent workflows. In other words, logic becomes the product.
For financial institutions, this raises uncomfortable questions. How quickly can core advisory processes be digitised and industrialised? What happens to large-scale transformation programmes when iteration cycles shrink from months to days? Where does differentiation move when implementation is no longer the constraint?
The pattern is unlikely to remain confined to finance. Any domain where expertise can be formalised — law, healthcare, engineering — may see a similar dynamic. The distance between idea and product compresses. The cost of experimentation falls. The speed of iteration increases.
This does not eliminate the need for expertise. It raises the bar for it. The challenge is no longer simply to build systems, but to structure, govern and control them.
And this is where the story that began at 2am becomes relevant again.
The speed at which that first prototype — and later, the 30-minute system — emerged was not a function of technical brilliance. It was a function of structured thinking. Two hours earlier, the idea existed only as a thought. Thirty minutes earlier, the second system did not exist at all. Shortly after, both were fully operational.
The difference was no longer the ability to code.
It was the ability to define.
AI will not be constrained by intelligence. It will be constrained by how well we structure thinking.
That may turn out to be the more profound shift.