The Articulation Gap

The other month I made a decision I’d normally document clearly — but didn’t.

We were working on a complex onboarding problem for data publishers. The workflows were intricate, the domain messy, and the complexity intrinsic. This wasn’t something to be simplified away. Our job was to help people navigate it, not pretend it wasn’t there.

We explored three approaches:

• A wizard
• A media library-style interface
• A timeline

In the meeting I said, “The wizard feels clearer.”

The product owner nodded.

We moved on.

Ten minutes, maybe.

A week later, when I tried to write the decision up, I realised how much reasoning had stayed in my head.

The wizard wasn’t just clearer. It allowed progressive disclosure of complexity - revealing information in stages, managing cognitive load, letting people focus on the decision immediately in front of them. The other approaches exposed the entire decision space up front. More flexible, perhaps. Also more overwhelming.

I know this. I’ve learned, sometimes the hard way, to articulate reasoning as I go because I’ve seen what happens when it disappears. But this time I slipped. We were moving fast. The choice felt obvious. I let it slide.

Articulation is fragile. Even with experience.

AI is making that fragility impossible to ignore.

Choices vs Decisions

A choice is:

“We’re going with the wizard.”

A decision is:

“We’re using progressive disclosure to manage intrinsic complexity in the publishing workflow, accepting slower task completion and higher development cost in exchange for reduced cognitive load and lower abandonment.”

Early in my practice I made mostly choices. I picked things that felt right and justified them later if asked. The justifications sounded like articulation, but they were really rationalisation - reasoning constructed after the fact to defend an intuitive call.

Real articulation happens before the choice. It’s the thinking that makes the choice possible.

The uncomfortable part is that it requires naming trade-offs explicitly. Which means admitting that someone is going to be unhappy. Designers avoid this not because they can’t articulate, but because articulation removes plausible deniability.

Vague reasoning feels safer.

Vague reasoning also doesn’t travel.

The Constraint That Quietly Disappeared

Producing competent design work is no longer the constraint.

I can open an AI tool, describe a user flow, and get something coherent back in ninety seconds. Decent hierarchy. Reasonable labels. Defensible interaction patterns. The work of a solid mid-level designer, on a good day.

The output isn’t the problem.

The output is fine.

What’s missing is the ability to say why this solution exists instead of another equally fine one.

AI can generate options endlessly. It cannot tell you which one matters, or why, or what choosing it commits your organisation to six months from now.

Judgment without articulation is just preference wearing fancy shoes.

That’s always been true. AI just makes it much harder to ignore.

What Articulation Actually Does

Articulation isn’t explanation.

It’s not documentation layered on after the real work.

It’s infrastructure.

It’s what allows decisions to survive contact with an organisation.

A well-articulated decision answers:

• What problem are we actually solving?
• What are we optimising for?
• What are we explicitly not optimising for?
• Who absorbs the cost?
• What would make us reverse this?

Without clear answers, decisions decay. Intent becomes interpretation. Interpretation becomes assumption. Assumption hardens into “we’ve always done it this way”.

The pattern is depressingly consistent:

• A good decision gets made
• The reasoning stays in someone’s head
• That person moves on
• Six months later the organisation is re-litigating the same ground because no one can reconstruct why the choice was made

When Reasoning Disappears

A few years ago I worked on a data portal for the United Nations. The core interaction was a map with project locations plotted on it.

There was a decision about whether to use a flat map or a 3D globe that smoothly converted to flat when you zoomed in.

The team chose the globe.

The reasoning was specific. Flat maps carry projection bias. Landmasses are distorted, centred, privileged. For a UN context, that bias mattered. The globe was more neutral.

Some time after launch, a senior stakeholder asked the obvious question.

Why the globe?

It was harder to build.

It performed worse on older devices.

Why not just use a standard flat map?

By then, most of the original team had moved on. I watched the remaining team struggle to reconstruct the reasoning.

“It’s more engaging.”

“Users interact with it more.”

Both true. Neither the reason.

The original articulation - projection bias, geographical neutrality, appropriateness for a UN context - had never made it into the documentation. It existed only in the heads of people who were no longer there.

The globe stayed. Not because the decision was successfully defended, but because it was “cool”, the metrics looked fine, and changing it would have been expensive.

The decision was sound.

The organisation learned nothing.

When the next map decision came up, they started from scratch.

When Volume Becomes The Problem

AI hasn’t just made production cheap. It’s made volume the default condition.

Yesterday I generated eight navigation structures in four minutes. All reasonable. None obviously wrong. Several genuinely interesting.

This is where less experienced designers get stuck. They can generate options endlessly, but they can’t select between them with confidence. Selection without articulated criteria is just taste.

I can often spot which option will work better because I’ve seen similar patterns play out before. But intuition doesn’t scale. It doesn’t teach. It doesn’t survive me leaving the room.

“This one feels clearer” helps no one.

“This one establishes hierarchy through spatial relationship rather than visual weight, which will scale better as we add features later” does.

When anyone can generate options, the value is entirely in selection and articulation.

What AI Is Actually Good At

AI is a pattern-completion engine. It extends what exists. It cannot decide what should exist.

I use it constantly. It expands half-formed ideas faster than I can. It explores structural variations I wouldn’t have thought of. It finds clearer phrasing for things I already understand.

But when I ask it which idea matters most, it gives me balance instead of judgment.

When I ask what to cut, it hedges.

When I ask which trade-off to make, it lists considerations without choosing.

That’s not a failure. It’s working as designed.

Leadership work happens under conditions AI cannot replicate — political constraint, incomplete information, organisational history, the quiet realities that never make it into a prompt.

AI can propose.

It cannot be accountable.

A Note On Method

I used AI to write portions of this article.

I sketched arguments. AI helped expand them, test phrasing, explore structure. A few sections are largely AI-shaped. Others are mine. Most are collaborative.

The boundary isn’t always clean. Some sentences that feel like my voice may be influenced by thousands of interactions with AI. Some AI-generated phrasing may be closer to what I actually think than my first draft was.

What remains entirely mine is the articulation - what to elevate, what to cut, which trade-offs to accept, which arguments will actually travel.

That’s not because I have a special human essence AI lacks. It’s because I have context - about organisations, incentives, failure modes, and how decisions tend to decay once they leave the room.

The method is part of the point.

What This Means For Design Leadership

I can articulate most of my decisions clearly now. I still catch myself slipping.

Making good intuitive calls without documenting the reasoning.

Letting obvious choices go unrecorded.

Assuming the context in my head will be obvious to others.

If this still requires effort, it suggests articulation isn’t something you master once and keep forever. It’s a discipline that needs maintenance.

AI makes this discipline more important and more difficult.

More important because output no longer signals judgment.

More difficult because AI can help you sound articulate without doing the thinking.

The work hasn’t changed:

• Looking at multiple defensible options
• Knowing which signals matter
• Making trade-offs deliberately rather than intuitively
• Articulating decisions clearly enough that teams can execute without you in every room
• Taking responsibility for what happens next

What’s changed is that you can’t hide behind craft anymore.

Only articulation travels.

Questions Worth Asking Before You Commit

These are the questions I keep coming back to — especially when a choice feels obvious.

Can you write the decision as a single sentence?

If it can’t travel, it isn’t ready.

Not:

“We’re going with the wizard.”

But:

“We’re using progressive disclosure to manage intrinsic complexity, accepting slower completion in exchange for reduced cognitive load and lower abandonment.”

If you can’t state it clearly, what you have isn’t alignment. It’s shared momentum.

What are you explicitly not optimising for?

Real decisions give something up.

If you haven’t named the loss, you haven’t made a decision yet. You’ve made a choice and convinced yourself it has no downside.

Who absorbs the cost?

• Users spend more time.
• Support handles more questions.
• Engineers maintain more complexity.
• The business accepts slower growth.

Most of the time, you know the answer. You just don’t want to say it out loud.

What signal would make you reverse this?

Good decisions include their own escape hatch.

What metric, behaviour, or feedback would tell you this isn’t working? When will you check?

If the answer is “we’ll know when we see it”, you probably won’t.

Can someone else explain this accurately without you present?

If another team can’t take this reasoning to a stakeholder and get it right, you’ve understood it privately. You haven’t articulated it.

What’s not in the design file?

AI can generate flows, diagrams, journey maps. Those are maps.

You’re navigating terrain:

• Organisational constraints
• Political realities
• Historical decisions nobody wants to reopen
• Regulatory exposure
• Harm that only appears at scale

That’s usually where the real decision lives.

And it’s the first thing to disappear when people move on.

The value of design leadership isn’t output.

It’s the ability to look at abundant options and know which signals matter. To decide deliberately. To articulate clearly enough that others can move with confidence.

Not a soft skill.

How organisations stay coherent when everything else is accelerating.

If You Enjoyed This Piece

Follow me for more articles about design-driven coding, usability, and the messy, wonderful middle ground between aesthetics and logic.

Author’s Note:

I am currently deeply interested in using AI to generate both visual and text-based content. I am actively collaborating with AI on multiple platforms to explore my thoughts on what creativity is and is not.

My current approach is to collaborate with AI by using the output as a foundation upon which to build and modify.