The Best Interface for an Expert Is a Substrate, Not an Assistant
A scaffold tells you what to do.
A substrate waits until you need it.


A novice needs a scaffold. An expert needs a substrate.
Design · UX · Enterprise
The myth of "helpful" design
Picture someone who's run the same workflow in the same tool for years enough that they don't think in screens anymore, they think in the work itself. Then an update ships. A dimmed screen. A pulsing beacon. A modal: "Welcome to the new workspace let us show you around."
To whoever's tracking onboarding completion, that modal is a success metric. To the person who's been doing this for a decade, it's an insult. It's the product saying: I don't recognize you, I don't know how long you've been here, and I'm going to treat you like it's your first day.
It's easy to read an expert's resistance to that kind of redesign as stubbornness "they just don't like change." I don't think that's accurate, or fair. What's actually being defended is hard-won automaticity: years of muscle memory, shortcuts, and assumptions about where things live, built up until using the tool stopped requiring conscious thought at all.
We keep treating an expert's reluctance to redesigns as resistance to change. It's closer to the defense of mastery they've already earned.
Ajaysinh Barad · Oct 05, 2026 · 7 min read
I spent the first part of my career designing games, and the years since designing enterprise systems most recently a product lifecycle and engineering tool used by people who've been doing their job for a decade or more.
I used to think of those as opposite disciplines. One holds attention. The other processes work. But living inside both taught me something I didn't expect: games solve the transition to mastery with real precision, and enterprise tools almost never do.
A beginner opening a game for the first time gets
hand-held, deliberately. An arrow, a prompt, a small win, repeated until the mechanic sticks. But the moment a player has internalized how the world works, a well-made game goes quiet. It stops interrupting. It trusts what it already taught you.
Enterprise software almost never learns to do
that second part.
Move a toolbar three pixels. Wrap a core action inside a new drawer. Add a tooltip that can't be dismissed for good. None of that reads as "cleaner" to that person. It reads as a small, repeated reintroduction of friction into something that used to be instinct.
Most products design one static state.
Expertise asks for a decay curve instead.


A novice needs a scaffold: visible, rigid, proactive. It tells you what to do, constrains your mistakes, and surfaces definitions you don't have yet.
An expert needs something closer to a substrate: silent, dense, passive. It exists underneath the work, offering nothing until force is applied to it.
The common failure in enterprise design is assuming the interface should stay identical across a user's entire relationship with the product as if day one and year ten call for the same amount of explanation. They don't. A tool that's actually paying attention to its user should behave differently as that user's fluency grows: fewer explanations, denser information, faster undo instead of upfront confirmation, shortcuts surfacing in place of walkthroughs.
None of that means expert tools need less design thinking. It means the design has to do something harder than add guidance it has to know when to retract it.
A scaffold, and a substrate
The quiet failure of "helpful" intelligence
This is the part that matters more every year, as more products add some version of an AI layer on top of the same workflows an assistant that predicts your next move, a suggestion that pops up mid-task, a copilot that wants to finish your sentence for you.
For someone new to a tool, that kind of help is a genuine gift. For someone who's already made the decision who already knows exactly what they're doing next the same behavior reads completely differently. It stops feeling like assistance and starts feeling like the system
second-guessing a judgment that didn't need checking.
Every suggestion carries a hidden cost: a half-second of evaluation. Is this right, is it worth considering, should I take it or ignore it. For a beginner, that half-second is cheap they didn't know what to do anyway. For an expert mid-flow, it's pure overhead, inserted into a decision that was already made.
Intelligence isn't just knowing when to act. It's knowing when not to.
The more capable these systems get, the more that restraint matters not less. An expert doesn't want a tool competing with their judgment. They want it to extend it, and extending something well usually means staying out of its way.
What this looks like in practice
Across the enterprise products I’ve led and contributed to, I’ve seen how a feature that looks helpful in a demo can feel very different in everyday use. What seems intuitive to someone encountering it for the first time can add unnecessary steps for someone who has worked with the same workflows for years.
The demo rewards the feature for explaining itself. The expert rewards it for staying out of the way.
A few things follow from that in practice.
Design onboarding for the people who need it. Don't add walkthroughs simply because a new surface needs one. Consider who is encountering it for the first time, how often that happens, and whether guidance is actually needed.
Give experienced users control. Hints, tips, and suggested actions can help people learn, but users should be able to dismiss them when they're no longer useful. Guidance should support the work, not compete with it.
Protect muscle memory. Moving a familiar control to accommodate a visual refresh may seem like a small change. For someone who has relied on that interaction for years, it can disrupt an established workflow. Preserve familiar patterns unless there's a meaningful reason to change them.
Measure what doesn't happen. Click-through rates, feature adoption, and session time can tell us something about product usage, but they don't necessarily tell us whether an expert can work without interruption. I'd also look at three things:
Uninterrupted flow how often people can complete their work without stopping to seek help they don't need.
Silent recovery how easily people can recover from mistakes without unnecessary alerts or support requests.
Unforced evaluation how often the interface asks people to assess suggestions or guidance they never needed in the first place.
These outcomes can be harder to measure than clicks or adoption. But they're worth considering when designing tools for people whose goal isn't to explore the interface. It's to get their work done.
Designing a boundary, not just a path
I don't think this makes expert-facing design less interesting than the consumer kind. I think it makes it harder, in a specific way.
Designing for a beginner is designing a path. Designing for an expert is designing a boundary deciding exactly where your help ends and someone else's competence begins, and having the discipline to actually stop there.
Most design writing doesn't spend much time on that skill, because most of it is written with the newcomer as the default reader, the same way most products are built with the newcomer as the default user. But the person in front of the screen isn't always new. Sometimes they've been there longer than the product has existed in its current form.
The simplest way I can put the distinction now: a beginner needs the product to show them a way in. An expert needs it to stay quiet enough that they forget it's there at all, and get back to the actual work.
The highest compliment an expert can pay a tool is that they stopped noticing it was there.


