Devesh JoshiCo-founder, product
Nine years building AI platforms serving 12,000+ engineers. LLM platforms, agentic systems (MCP), and multi-model safety evaluation.
Design debt: why UX bolted on late costs more
A design decision is cheap in week one and expensive in week nine. The cost curve, and the three decisions that cause most of it.
The cost curve is not linear
Moving a wall costs nothing while it is a pencil line, a little while it is a flat, and a fortune once it has been lit and shot. Interface decisions behave the same way.
Changing a navigation model during design costs an afternoon. Changing it after the routing, state shape and permissions have been built around it costs a sprint and introduces bugs in places nobody associates with navigation.
Three decisions that generate most of the debt
The navigation model. It determines the route structure, which determines where state lives. Decide it late and you are refactoring data flow to accommodate a menu.
The empty and error states. Teams design the happy path, build it, then discover that half the real screens are neither empty nor happy. Retrofitting those states means touching every component that can fail.
Whether it is a design system or a set of screens. Screens are faster for the first five and slower for every one after. The crossover arrives sooner than most teams expect.
What design in week one actually looks like
Not a full visual language. It looks like deciding the navigation model, naming the states every screen can be in, and agreeing the type and spacing scale — three artefacts that take days and constrain thousands of later decisions.
This is why design sits inside Discover and Design rather than arriving after Build. It is not a preference about process. It is where the decisions are cheap.
If the product is already built
Do not start with a redesign. Instrument it, find the two screens where users actually stall, and fix those. A full redesign of a product with a known drop-off point is usually an expensive way to avoid reading the analytics.
Topic Focus & Target Concepts
This is the work behind our design practice — UX, design systems, and motion built into the engagement from day one rather than bolted on after the architecture is set.
Product and interface designRunning into this in your own stack? Twenty minutes, no deck.
Book the call