Vibe-coded tools are building on top of the advisor stack

Every few years, the advisor technology stack starts to look dangerously self-buildable. The latest candidate is everything: AI tools will turn plain-English instructions into code and, the theory goes, make third-party software less necessary. A new analysis on Kitces' Nerd's Eye View argues that 'vibe-coded' software is not about to hollow out the AdvisorTech market. The build-versus-buy math, it says, has moved far less than the hype suggests.

The old math was unforgiving. A third-party provider spreads development costs across many users; a firm that builds its own tool pays for the whole thing alone. The Nerd's Eye View post says a custom build can take years to realize enough savings to recoup the initial investment, by which point the software is often due for a replacement cycle. For most firms, buying was the rational default.

AI-powered vibe coding lowered the front-end cost. Instead of paying a developer to type out every module, an advisor can describe what they want in plain English and get code back. That changed the entry threshold, the post says. Building once made sense mainly for firms with 20 or more advisors; smaller firms and solo advisors can now take a swing at custom technology. The shift produced a familiar set of predictions: mass license cancellations, consolidation of the AdvisorTech market, a homegrown software revolution.

The early evidence points elsewhere. Advisors who are all-in on tools like Claude are mostly building on top of the vendors they already use, the post reports, not around them. In the Builder FP community, where these conversations concentrate, two kinds of projects dominate. One is integrations that pull reporting outputs from existing platforms into a single client deliverable. The other is niche tools built for a specific type of clientele.

That split makes sense. Most advisors are fairly happy with the software they already use, the post notes, and there is little reason to rebuild what works. The system of record, the billing, the connection to custodians — the parts an RIA would want to replace — carry years of accumulated requirements and a compliance trail. A vibe-coded replacement starts from zero on all of that, and the maintenance burden falls on the one firm that needs it. The additions a niche practice wants are small, specific and close to its own workflow.

The post draws a more contrarian conclusion about supply. Instead of shrinking the third-party market, the spread of AI coding tools will likely increase the number of solutions available. The reasoning follows from the same dynamic that made building cheap: a tool created to solve a single practice's problem is a prototype for a product a vendor can refine and sell to a hundred firms. Every practice that builds for its own niche has just produced a feature list for someone else to ship.

The practical test for a practice is not build versus buy. It is build the edges, buy the core. Three questions sharpen it. Does the workflow recur often enough that a few saved hours each week add up? Is the requirement so specific that a mainstream vendor will never build it well? Can the firm tolerate maintaining the tool itself, including the day the AI that generated it starts acting differently? If the first two are yes and the third is honest, build. If not, the license fee is still the cheaper answer.

The firms that get the most from vibe coding will be the ones that understand the distinction. They will not be describing a rip-and-replace project in plain English. They will hand Claude a narrow problem, put the result alongside the systems that already work, and skip the part where they become a software company.