The argument

The future of B2B SaaS is buy the system, bring your own UI.

AI changes the total-cost-of-ownership calculus. It makes more interfaces economical to build; it does not make the underlying systems disposable.

AI coding has made it possible to prototype in days what once took quarters. That is a meaningful economic shift. But it does not eliminate the build-versus-buy question; it sharpens it.

The smartest companies have always built what is closest to their differentiation and bought infrastructure that would be risky, expensive, and distracting to recreate. The new opportunity is to split the decision into layers.

What remains expensive

The hard part of mature SaaS is rarely drawing the chart. It is harmonizing systems, maintaining integrations, resolving identity, enforcing permissions, preserving history, handling edge cases, modeling the domain, and making actions reliable.

When the cost of front-end code falls, the value of a trustworthy foundation can rise. A custom interface on the wrong model simply spreads confusion faster.

What becomes fluid

The native product remains important as the default, fully supported experience. It simply stops being the only possible surface for every user, question, and moment.

A demand-generation leader may want a board-ready narrative. A sales leader may want an account-priority queue. A marketing-operations leader may need a lineage and anomaly debugger. A content team may need a workspace spanning beehiiv, HeyGen, and distribution channels. Those interfaces can be narrow because the systems underneath remain broad.

Seven tenets

  1. The interface is separable. A SaaS application bundles system and experience, but the two do not have to remain inseparable.
  2. The system is still valuable. Governed data, semantics, actions, permissions, and reliability are durable product value.
  3. The native UI still matters. BYO-UI extends a strong default product; it is not an argument for abandoning one.
  4. Interfaces should follow moments of work. Build around a decision, exception, approval, or production task—not a clone of the source application.
  5. Use several access modes. SQL, APIs, events, MCP, SDKs, and extensions are complementary rather than mutually exclusive.
  6. Governance travels with capability. Identity, scopes, approvals, audit, provenance, and rollback are part of the interface contract.
  7. Composition is first-class. A customer-owned interface can sit over one platform or coordinate several without erasing their sources of truth.

What BYO-UI is not

  • It is not “vibe-code your own Salesforce.”
  • It is not a claim that every workflow belongs in a chat window.
  • It is not permission to bypass consent, policy, or approval controls.
  • It is not an MCP-server marketplace.
  • It is not a product, service, or vendor ranking.

Claims worth testing

The thesis should be judged by observable outcomes: whether customers adopt systems more readily when interface lock-in falls; whether vendors can expand into more personas without bloating the native product; whether focused interfaces improve decision time and adoption; and whether governance can remain intact as experiences become more fluid.

The practical test: choose one persona, one decision, one trusted system, and one narrow interface. Begin read-only. Show source evidence. Add one reversible or approval-gated action only after the value is clear.

Original formulation

The thesis was first published by Harry Hawk in April 2026 and then extended through public discussion and this research Atlas.