The Old Habit

The old habit was comforting: kickoff, then frames, then components, then a prototype that looked finished enough to calm everyone down.

Stakeholders could react to colors and cards. Engineers could estimate screens. The file felt like progress.

Too often we had designed a plausible interface for a fuzzy workflow. Beautiful wrong is still wrong — and expensive to unwind once people fall in love with the mock.

What Pixels Hide

High fidelity hides uncertainty under polish. Gaps in the process look like missing icons instead of missing decisions.

Debate moves to button radius while the real questions stay soft: Who owns this step? What data is required? What happens on failure? What is the exception path that eats half the week?

Pixels are concrete. Ambiguity hates concrete objects. So the team argues about the concrete thing and postpones the ambiguous one — until build time, when ambiguity becomes change requests.

Delay high-fidelity until the problem is high-clarity.

Start With the Work

I start with workflows on paper, in a doc, or as a ruthless outline: trigger, decisions, outputs, handoffs, and failure modes.

No chrome. No type scale. Just the path a human takes to finish a job. If we cannot narrate that path cleanly, we are not ready for frames.

This feels slower on day one. It is faster by week three, when the UI is assembling a known process instead of inventing one through layout debates.

Users and Data Before Layout

Next come two boring questions:

• Who is acting at each step, with what permissions and motives?
• What data must exist, be trusted, and be editable here?

Screens are containers for answers to those questions. If the answers are mushy, the containers will be pretty and hollow.

I have watched teams design elegant empty states for data models that could not support the happy path. The Figma file was not the problem. The sequence was.

When Figma Earns the Front Seat

Figma earns the front seat when the workflow is stable enough that visual exploration changes the quality of the experience — density, hierarchy, interaction nuance, brand expression.

It also earns it for alignment: showing a sticky interaction pattern across a suite, stress-testing a complicated state matrix, or giving engineering a precise target once the product decisions are no longer moving weekly.

Tools are not the enemy. Premature fidelity is.

A Better Sequence

Sometimes step four is a whiteboard photo. Sometimes it is a rough HTML skeleton. The medium matters less than refusing to skip the thinking the medium cannot do.

Low fidelity is not anti-design. It is design that still allows the truth to move. Boxes and verbs invite correction. Polished frames invite attachment. Attachment is expensive when the workflow is still wrong.

Handling Stakeholder Anxiety

People get nervous when they cannot see pretty screens yet. Anxiety often sounds like urgency: “Can we just see the UI?”

I answer with artifacts that are still concrete: a workflow map, a decision table, a list of states, a one-page narrative of a user finishing the job. Those are reviewable. They just do not pretend the hard questions are already settled.

If someone needs visual reassurance, offer a single key screen in rough form — structure only — after the workflow is agreed. Resist the full kit of polished pages until the path holds up under questions.

The goal is not to withhold design. The goal is to stop paying for redraws that were really undecided product choices wearing a UI costume.

What Changed in Practice

Projects feel calmer. Stakeholders argue about the work earlier, when changing direction is cheap. Engineers get fewer surprises dressed as “small UI tweaks.” Design time goes into the moments that deserve craft.

I also enjoy Figma more this way. It becomes a place for craft instead of a dumping ground for unresolved product anxiety. The file gets thinner. The decisions get sharper. The build matches the intent more often.

I no longer start projects in Figma. I start with workflows, users, and data. Pixels come after the work is understood. UI is still essential — it just stopped pretending to be the first understanding.