The Best I Could Build in a Week

A fully featured advisor desktop front end — household views, portfolio analytics, AI-assisted workflows, collaborative task management — built from a blank directory to a clickable prototype in seven days. Not by a team. By one product leader and an AI pair programmer.

That is not a flex. It is a data point.

I argued recently that wealth management technology operates on pace layers — the front end moves fast, the infrastructure moves slow, and the gap between them is where the industry’s real cost lives. The five-month, five-million-dollar button. This piece is the proof of that argument from the builder’s side: what does the fast layer look like when you strip away every slow-layer constraint?

Opinions, not features

The prototype is not a feature demo. It’s an argument about what advisor software should be — expressed in code instead of slides.

Most advisor desktops treat collaboration as an afterthought. An advisor, a planner, a service associate, and a portfolio manager all touch the same client — but the tools make them work in isolation. This one started from the opposite premise: servicing is multiplayer. Embedded chat, shared task cards, and a unified activity feed made the team’s work visible by default, not buried in separate modules.

AI showed up the same way a colleague does. The agent interface lived inside the same messaging surface as the human team — not in a separate chatbot window, not behind a “magic wand” icon. Assigning a rebalancing review to an AI agent used the same workflow as assigning it to a junior analyst. The distinction between human and machine work faded into the background, which is where it belongs.

The portfolio view handled global multi-asset-class households natively — equities, fixed income, private assets, digital assets, worldwide currencies — because the client base has changed faster than the tools have. Challenger fintechs like Coinbase and Robinhood are collapsing these silos for retail investors. The wealth management industry is still asking clients to view their lives through product-line lenses.

Navigation moved from household to client to account to individual sleeve without friction. No page reloads, no context loss, no “go back to the dashboard” dead ends. This sounds like table stakes. In most enterprise wealth platforms, it is not.

How it was built

I spent the first day co-creating a detailed product requirements document and design spec with Claude — drawing on fifteen years of opinions about what wealth management software gets wrong. The PRD captured product decisions, interaction patterns, data models, and the specific tradeoffs I wanted the prototype to make.

From there, Claude Code built against the spec. I loaded domain-specific skills — financial modeling conventions, portfolio analytics patterns, custodial data structures — alongside development workflow skills that enforced planning, architecture review, and iterative feedback loops. The AI did not replace product judgment. It amplified fifteen years of it.

I gave active design feedback, asked questions about engineering tradeoffs in planning mode, and explored options together. I have limited ability to evaluate whether the underlying code is elegant or monstrous — but it works, it feels smooth, and for a prototype, that is the only test that matters.

What the prototype could not touch

The prototype ran on clean mock data I generated. Not production data reconciled across three custodians, two portfolio accounting systems, and a CRM that was last audited in 2019. It had no entitlements model — no security review asking who can see which client’s alternative asset holdings at which level of aggregation. No legacy API with a thirty-second response time that turns a real-time dashboard into a loading spinner. No competing stakeholder opinions about whether the AI agent should live in the sidebar or get its own tab.

These are not flaws in the prototype. They are the point of it.

Every constraint I removed is one that takes months to navigate in production. The front end is solved — not just technically, but as a product problem. The design thinking exists. The tools exist. What is missing is the infrastructure beneath — the data foundations, the entitlements architecture, the system-of-record performance — that determines whether any of these ideas can ship.

The fast layer, proven

A week for the prototype. Months and years for the infrastructure to make it real.

The gap is not a technology problem. It is a structural one — and it is where the next phase of competitive advantage lives. The firms doing the unglamorous work beneath the surface, cleaning data, rationalizing post-acquisition architecture, building entitlements models that move at the speed the front end demands, will be the firms who attract, retain, and arm advisors with the tools they need to compete and win.

The prototype is the easy part.