What vibe coding taught me about enterprise UX

1. How it actually started

The story usually told about Zero UI is that I saw the future. The actual story is that I had a productivity problem.

The HONO engineering team was scaling fast. The new-joiner ramp was uneven. The output quality from senior engineers and junior engineers was further apart than the org chart suggested. I introduced AI pair programming to the team — Microsoft Copilot first — to close that gap. The framing I used at the time was deliberately small. Every engineer should have a senior pair partner sitting next to them, even when there is no senior engineer available. The output would converge on a standard. The juniors would ramp faster.

That was the pitch. It worked the way I expected, for the reasons I expected, for about a year.

What happened after that was not on my roadmap.

2. What changed without me directing it

Some of the engineers, out of their own curiosity, started using Cursor instead. I had not mandated it. They liked the chat-pane shape better than the inline suggestions Copilot was giving them. They kept shipping. Output kept improving. I noted the divergence and moved on.

Over the next several months, the Cursor adoption spread organically across the team. New joiners would arrive, see the engineers around them working in a chat pane, and pick up the same workflow. The tooling choice was no longer something I needed to decide. The team was deciding it.

By late 2025 the inflection had completed. The team was fully on Claude as the primary pair programmer. The original framing — AI as a sitting-next-to-you pair — had quietly become something else. AI as the surface they were working through.

I had introduced this evolution. The shape it had reached was not the shape I had introduced.

3. What actually changed

The change was not “engineers work faster.” The change was that the part of engineering work that used to be navigation had been replaced by articulation.

A senior engineer in 2022 would normally spend a morning the same way they had for a decade. Read the docs of a library. Open a few tabs to understand the API surface. Write a small spike. Delete the spike. Write the real thing. Productive, but click-heavy. Read, scroll, copy, paste, run, fix, repeat.

The same engineer in late 2025 spent the morning describing what they wanted to a model, reading the model's draft, refining it, asking the model to rewrite a section, accepting a final version. The clicks were replaced by sentences. The reading still happened, but not docs. Code.

The afternoon I walked across the engineering floor and noticed nobody was typing into the editor — the typewriter rhythm of programming, the one I had been listening to in engineering rooms for fifteen years, gone — was the afternoon I knew the inflection had reached the room.

Articulation had replaced navigation. In the most click-loyal workforce I had ever managed.

4. The question I couldn't shake

The thing about engineers is that they are the most click-loyal users any software company has. They learn keyboard shortcuts. They customise their IDEs for a decade. They develop muscle memory for build commands and test runners. If anyone was supposed to keep clicking, it was them.

And they had stopped.

I started asking the question out loud in conversations with peers, and then in my own head at three in the morning.

If developers — the most click-loyal users in the company — won't click anymore, why should HR users?

HR users had been forced into click workflows for thirty years. Not because they wanted to navigate four menu layers to apply for leave or reconcile a payroll line or find a contractor's timesheet. Because the alternative did not exist. The cost of articulation, of turning intent into action without intermediate UI, was higher than the cost of building the menus.

Now it was not.

I could see exactly where this went. And I could see that the HR-tech industry, including the company I had just joined, was not yet looking at it.

5. What Zero UI actually is

The Headless HRMS, the system that became HONO Zero UI, is not a chatbot bolted onto an HRMS. It is not a copilot sitting alongside the existing screens. It is the same shift that happened to my engineers, generalised to everyone in the enterprise who needs to do work in software.

A line manager wants to know which of his team's performance reviews are due next month. Old shape: log into the HRMS, navigate to performance, filter by team, sort by date. New shape: ask. The system understands the intent, fires the right APIs, and renders the result inline.

A finance partner wants to reconcile this month's payroll variance against the budget line. Old shape: export to Excel, pivot, compare. New shape: ask. Same execution layer, different intent surface.

An HR business partner wants to know which of her attrition-risk employees have skipped their last two one-on-ones. Old shape: there is no screen for that question; she would build the answer by hand from several reports. New shape: ask. The system composes the answer from the tools available to it.

The screens did not vanish entirely. The screens stopped being the workflow. They became the artefact the system chose to render — sometimes a list, sometimes a chart, sometimes a card, sometimes nothing at all if the answer was a sentence.

6. What this means for enterprise UX more broadly

The screen-and-click UX paradigm is not going to vanish in a single year. Excel will outlive me. Figma will outlive Excel. Some interactions genuinely need visuals or direct manipulation, and they will keep their screens for very good reasons.

But the vast middle of enterprise software, the workflows that exist because someone built a screen for them rather than because the screen is the point, is going to be unbundled. Form-driven configuration. Filter-and-find. Multi-step approval. The thirty-year backlog of “first click here, then click here.” All of it compresses into one question and one answer.

The companies that win the next decade of enterprise software are not the ones that bolt a chat surface onto their existing product. They are the ones that build for the new shape and let the old shape be a fallback when it is actually the right tool.

The mistake will be subtle and expensive. A vendor will ship a chat widget that looks like the conversational surface but that, underneath, is just running the same old workflow through a new front door. The user will still wait for the same screen to load. The user will still see the same five-step wizard, only narrated by a model. That is not the shift. That is a chat skin on the old shape.

The shift is the execution layer. The shift is what happens when the system stops being a tree of pages and becomes a vocabulary of intents that the model can compose into action.

7. What it doesn't mean

I am not predicting the end of UI. Some interactions genuinely benefit from visuals: a calendar, a Gantt chart, a chart of headcount by region over time. The user does not type their way into those. They look at them and they interact with them directly.

What changes is the default. The default surface is no longer “here are all the screens, find the one you need.” The default surface is “tell me what you need, and I will render exactly what answers it.”

Inline generative components, slash commands, voice input, on-screen context as input — these are still interfaces. They are just composed dynamically by the execution layer rather than pre-built and waiting for the user to navigate to them.

The interface is not the product. The execution layer is.

8. The engineers showed us first

Engineers did not make a tool choice in mid-2024. They demonstrated, in the most click-loyal workforce in the company, that articulation had become cheaper than navigation. That demonstration was the signal.

The rest of the enterprise will follow. Not all at once. Not in every workflow. But the direction is set, and the companies that build for the destination instead of renovating the road behind them will be the ones the next generation of users actually choose.

For HONO, that is what Zero UI is. For everyone else, the question is the same one I asked myself walking across that engineering floor.

If your most click-loyal users have stopped clicking, what are you still building screens for?