A position on product design
Design is not UI.
Prompt is not strategy.
UI is what a product looks like. Design is the judgment about what it should be, and why.
AI has transformed how we build. It accelerates prototyping, explores variations instantly, and handles repetitive work that once consumed entire days. It is genuinely useful.
And yet, something important is getting lost in the excitement.
Scroll to understand why this distinction matters.
•5 min read
What AI does brilliantly
- •Creates visual options in seconds
- •Explores design variations at scale
- •Handles repetitive tasks
- •Speeds up the path from idea to screen
- •Lowers the cost of trying ten ideas instead of one
We use these tools every day. They have made our cycles shorter and our exploration wider.
What follows isn't a critique of AI. It's a closer look at what these powerful tools cannot replace. And why that matters for your product.
This is the biggest shift in design tools since the move to the browser. The teams that win are the ones who treat it as leverage, not as a substitute for thinking.
UI is the visible surface of decisions made one, two, and three layers above it.
The core misunderstanding
Common beliefs
Design = screens
Visual outputs as the goal
UX = UI
Mixing up experience with interface
Strategy = good prompts
Letting tools do the thinking
AI output = correct output
Accepting suggestions without question
Reality
Design = decision-making
Choosing what to build and why, based on evidence
UX = systems and behavior
Measured through real user actions
Strategy = trade-offs and intent
Weighing evidence with purpose
AI output = a guess
Something to test against real data
Fluent output reads as correct output. The more confident a result looks, the less anyone checks it, which is exactly when judgment matters most.
The left column can be automated. The right column is where products are won or lost.
UX is not UI.
Understanding the distinction
UI — User Interface
UI is an output: a surface, a skin. It's what users see and touch. It can be templated, generated, and styled from existing patterns.
Buttons, colors, layouts, typography, spacing, icons.
UX — User Experience
UX is a system of decisions. The logic beneath the surface. It determines what exists, why it exists, and how it behaves when things go wrong.
Information architecture, user flows, error handling, edge cases, mental models, cognitive load.
A polished interface routinely hides a broken experience. The surface tells you nothing about the structure beneath.
When organizations mix up UI with UX, they focus on appearance while the underlying experience suffers. Users feel this immediately, even when they can't explain it.
Consistency isn't about visual sameness. It's about coherent thinking: the same logic, the same care, the same understanding applied everywhere users interact with your product.
You can automate UI.
You cannot automate the judgment that decides what to build.
The design process
Research
Gathering data about the problem space
↓Synthesis
Finding patterns in evidence
↓Decisions
Choosing directions informed by data
↓Validation
Testing assumptions with real users
↓Outcomes
Measuring impact against success criteria
Each stage requires human judgment backed by data. AI can help with outputs at any stage, but the decisions between stages, reading the evidence, applying clear intent, and measuring results, remain human work.
A prompt can produce any of the dots. Strategy is knowing which one is worth picking.
Prompting is execution.
Strategy is intent.
Understanding the difference
A prompt is an instruction. Strategy is knowing which instructions are worth giving in the first place.
AI-first teams often confuse better prompts with better thinking. But a more sophisticated prompt cannot compensate for unclear goals, untested assumptions, or missing data about user behavior.
A better prompt cannot fix a missing strategy.
Prompts do not define goals.
They execute within given constraints.
Prompts do not weigh trade-offs.
They optimize for specified outputs.
Prompts do not know your users.
They pattern-match from training data, not lived experience.
Prompts do not maintain consistency.
Each response is stateless, without memory of your product journey.
Strategy requires understanding what success looks like before you begin. No prompt can provide that for you.
Decide sits upstream of everything. The other stages execute against it.
What strategy actually decides.
The calls a prompt cannot make for you
Strategy is not a longer prompt or a smarter model. It is a small set of decisions made before any work begins.
These decisions are mostly subtraction. They narrow the field so the team can move with intent. Prompts can only add.
What problem is worth solving this quarter.
Out of every possible direction, you pick one. That choice is upstream of every prompt anyone will ever write.
Which users to disappoint on purpose.
You cannot serve everyone equally. Choosing who matters most is a judgment call, not an output.
What to cut so the rest can be great.
Most product work is subtraction. AI is built to add. The two pull in opposite directions.
How success will be measured before work begins.
If you cannot say what good looks like up front, no amount of execution will tell you when you have arrived.
AI has no accountability.
Designers do.
The accountability gap
When a product fails, when users churn, when support costs spike, when the market rejects your solution, someone needs to diagnose why.
AI cannot perform this function. People with product judgment can. They've seen patterns across products, felt the weight of failed launches, and learned what the data means in context.
Strategic judgment is earned, not installed.
Polished execution used to be the moat. AI commoditized that overnight. What's left to defend is the judgment behind the decision: which problem to solve, who to disappoint, and what "good" looks like before the work begins.
A more capable model does not close this gap. The constraint was never how good the output is. It is that judgment carries accountability, and accountability has to belong to someone.
Accountability cannot be automated. Someone still has to answer for the decision when it lands in front of users.
Bad design decisions compound.
Understanding design debt
Design debt works like technical debt. Every shortcut, every "we'll fix it later," every decision made without data adds up over time.
The cost shows up in the metrics: support tickets pile up, churn increases quietly, time-on-task grows, satisfaction drops. Teams spend cycles on rework instead of new value.
Users don't forgive inconsistency. They leave.
Consistency isn't about rigid patterns. It's about clear human intent across every interaction. Only people who understand the whole can make the parts work together.
How design debt accumulates
Each row was a decision that felt fast at the time. Together they become the product.
Same team, same calendar. The work that thinks first compounds; the work that ships first repeats.
The cost of skipping strategy.
Two timelines, same six months
Without strategy
- −First version ships in days.
- −Second version is a rebuild, not a refinement.
- −Support tickets cluster around flows nobody validated.
- −Churn rises quietly while the team ships features.
With strategy
- +First version ships a little later.
- +Each release builds on the last instead of replacing it.
- +Support volume stays flat as the product grows.
- +Retention compounds because the foundation holds.
Speed without strategy is just faster rework. The team that thinks first ships less, sooner, and keeps shipping.
None of this is news to you. You've felt the tension between what you know the product needs and what the timeline allows, between evidence and assumption, between craft and expedience. The question was never whether these problems exist. The question is how to talk about them with people who haven't felt them.
Shifting the conversation.
From aesthetics to outcomes
Evidence over opinion.
Usability recordings of real users struggling with "simple" flows accomplish what no slide deck can. When design decisions are backed by measurable impact on conversion, satisfaction, and time-to-task, the conversation shifts from taste to evidence. Fixing usability in production costs ten times more than getting it right in design, and that math speaks for itself.
Partner, not order-taker.
Product managers own the why: business goals, customer problems, what to build. Designers own the how: interface, interactions, user journey. When UX is involved in discovery rather than just execution, the inputs improve. Better inputs, better outputs.
Edge cases are where products break.
Stakeholders focus on the happy path. Designers specialize in what happens when things go wrong. When given a prescribed solution, returning with two or three alternatives that solve the same problem reframes the relationship. Edge cases ignored today become support tickets filed tomorrow.
Internal voices are not enough.
Industry examples of design investment paying off carry weight that internal advocacy cannot. Competitors with better UX did not arrive there by accident. Design systems reduce stakeholder burden by creating a shared language for interactions, turning repeated debates into resolved decisions.
Phrases that shift the conversation
When design consistently delivers measurable business results, stakeholders stop prescribing solutions and start relying on expertise.
What this site is (and isn't)
This is not anti-AI.
AI is evolving fast, and we welcome it. These tools make our work smoother: faster prototyping, wider exploration, fewer boring tasks eating up our days.
We're making the case for involving people with real product judgment, not anyone whose claim to design ends at access to the tools. Tenure is not the line. Strategic thinking is.
This is pro-expertise.
Years of practice, failed experiments, measured results, and deep user understanding cannot be prompted. Confusing tool access with real expertise is where products go wrong.
This site is:
- +Pro intelligent use of tools
- +Pro data-informed decision-making
- +Pro design as a measurable discipline
- +Pro accountability in product decisions
This site is not:
- −Anti-automation
- −Anti-efficiency
- −Anti-progress
AI amplifies thinking. It does not replace it.
AI accelerates Evidence and Execution. Goals, Decisions, and Measurement stay with you.
Where AI actually belongs in the strategy loop.
A simple map
Strategy is a loop, not a step. AI is excellent at one stage and useful in another. The other three are still yours.
01
Goals
Human only
Deciding what to pursue and what to ignore.
02
Evidence
AI assists
Synthesis, summarization, pattern surfacing on data humans collected.
03
Decisions
Human only
Weighing trade-offs against context AI cannot see.
04
Execution
AI shines
Producing artifacts, variations, prototypes, and first drafts at speed.
05
Measurement
Human only
Judging whether the outcome served the goal you set.
When teams let AI drift into Goals, Decisions, or Measurement, the loop breaks. The product starts optimizing for what is easy to generate instead of what is worth building.
The trap is that this drift feels like progress. Every stage you hand off removes friction today, so the speed is real. But the friction you removed was judgment, and its absence compounds: the product gets quicker to produce and slower to trust.
There is a harder version of this worth sitting with. The claim, made by people who take it seriously, is that AI is less a tool you pick up at chosen stages than an environment that shapes how you read the problem, coloring even the decisions you meant to keep for yourself. It is one perspective, and easy to overstate. But it lands on something true: being present in the loop is not the same as exercising judgment in it. The human stages have to be worked, not just occupied.
None of this runs against where AI is heading. The most advanced products are built this way on purpose: the human stays on the decisions, and automation expands only as it earns trust.
In closing
Design is how strategy becomes reality.
Every product is the sum of decisions made visible to a user. Some were made consciously, by people who understood the trade-offs. Others were made by default, by tools, by rushing, by not asking questions.
Users experience both. They just can't tell you which is which.
Experience still matters.
AI is a valuable tool and we use it daily. Judgment is not a tool. It is earned over years of shipping work and watching how it lands.
If this resonated, share it with someone who needs to read it.
Further reading
- Google PAIR — People + AI Guidebook on supervising automation and widening it only as trust grows.
- Microsoft HAX Toolkit — Guidelines for Human-AI Interaction on calibrating trust and narrowing an AI's scope when it is uncertain.
- Microsoft Design — UX Design for Agents on transparency and human control as foundations for AI agents.