The 5-Month and $5M Button

Stewart Brand’s pace layers model describes how complex systems change at different speeds. Fashion reinvents itself every season. Commerce cycles in years. Infrastructure shifts over decades. The fast layers innovate. The slow layers stabilize; they constrain the fast ones. No amount of speed at the surface changes what the foundation can support.

Wealth management technology runs on the same principle. The front end — the advisor dashboard, the client portal, the reporting view — is the fast layer. A skilled engineer can build a new feature in a day. With modern AI tooling, a working prototype takes hours. The surface has never moved faster.

But adding that same feature in production takes five months and costs five million dollars.

The button is not the problem. Everything beneath it is.

The pressure test

I built a prototype recently — a fully featured advisor desktop prototype, from product spec to clickable demo, in a week (~10 hours and <$100 on tokens). I wrote a detailed PRD drawing on fifteen years in wealth management, generated clean mock data, and built against a modern stack with zero legacy constraints. The result looks on par with proof of concepts I’ve seen from major software vendors and consulting organizations.

But the constraints prove the point — the front end is the “easy” part. I mocked up everything that’s hard, that holds back actual change in production.

The layers below the button

The data layer. The button needs data. Is the API available from the system of record? Is the data clean? Is it consistent across middle- and back-office systems? In most firms, the honest answer is: partially, from three different sources, reconciled on a spreadsheet someone built in 2014. The front end can be beautiful. The data behind it determines whether it is trustworthy.

The security layer. Every new surface creates a new entitlements question. Who can see this data? At what level of aggregation? Can the permissioning model enforce those rules at the speed the front end demands, or does every new feature trigger a six-week security review? Data governance is not a checkbox. It is a structural constraint that shapes what the front end is allowed to do.

The architecture layer. The system of record needs to be reliable and performant enough to serve real-time front-end experiences without minutes of loading spinners. In many firms, the back office was built for batch processing — not for the sub-second response times that modern interfaces demand. The plumbing was never designed for what the front end now asks of it.

Each layer constrains the one above. Dirty data means the API cannot be trusted. A brittle system of record means the API cannot perform. A rigid entitlements model means the feature cannot ship. The five months and five million dollars are not spent building the button. They are spent navigating the layers that determine whether the button can exist at all.

Two of everything

This is not only a problem for large incumbents sitting on decades of legacy architecture. A record wave of M&A across every firm size — RIAs, broker-dealers, hybrid firms, custodians — means that even fast-growing firms inherit what amounts to two of everything. Two custodians. Two portfolio accounting systems. Two client portals. Two CRMs with overlapping but inconsistent data.

Every acquisition multiplies the integration surface and deepens the infrastructure debt. The firm that closed three deals in eighteen months does not have a clean technology stack. It has three stacks duct-taped together, and every new button has to work across all of them. Growth, paradoxically, makes the slow layers slower.

AI widens the gap

The instinct is that “AI solves for this”. That large language models and intelligent automation will paper over the messy data, bridge the inconsistent systems, and let the front end sprint ahead regardless of what sits beneath it.

AI accelerates the fast layers. It makes the front end faster, the prototypes more impressive, the distance between “what’s possible” and “what’s in production” more visible. But it does not clean the last 10% of data foundations. It does not rationalize three overlapping custodial integrations. It does not resolve the entitlements model designed for a firm half the current size. Not this year anyway.

What AI does is widen the gap between what the surface could do and what the infrastructure allows. The demo gets better. The production deployment stays painful.

Where the advantage compounds

The next phase of competitive advantage in wealth management does not belong to the firms with the most impressive AI demos or the slickest advisor dashboards. It belongs to the firms doing the unglamorous, structurally consequential work beneath: cleaning data foundations, rationalizing architecture post-acquisition, building infrastructure that lets the fast layers actually move.

The firms that treat tech debt as revenue debt — because every month that button cannot ship is a month of unrealized advisor productivity, degraded client experience, and margin left on the table — will compound an advantage.

The fast layers get a follow up meeting. The slow layers win the deal.