Skip to main content
UX Design Engineering
Settings

Settings

Theme
Text size
100%
Typeface

Chapter 11

Developing skills and using AI with judgment

Choose a learning task from real work, get useful feedback, and verify the assumptions and behavior of AI-assisted artifacts.

After the assignment project, you have a list of things to learn: research, typography, TypeScript, testing, service design, and stakeholder communication. Starting six courses would feel productive. It would also make it difficult to tell which learning changes your work.

Choose the next skill from a concrete difficulty. The decision is what to practice, where to get feedback, and how to demonstrate that you can apply it again.

Diagnose the gap from an incident

Describe a moment when progress depended on help or a decision proved weak. Maya could build the dialog but needed support distinguishing component behavior from service responsibility. Jordan initially selected bulk assignment before understanding eligibility. Avery assumed that better documentation would solve adoption when teams also needed funded migration capacity.

These are fictional examples of different learning needs. They are not evidence that every junior, senior, or staff practitioner has the same gap. A staff engineer may need visual-design practice; a junior practitioner may already have strong research skills.

Record the task, what you tried, the feedback received, and what you could not yet do independently. That account is more useful than describing yourself as “bad at strategy.”

Choose one bounded practice cycle

A practice cycle contains a target behavior, a small task, feedback, a revision, and a second application. For visual hierarchy, compare two layouts with the same data and ask a designer to examine the task order. For uncertain outcomes, implement and review a lost-response scenario, then apply the reasoning to a different mutation.

The second application matters. Following a corrected example can show that you can reproduce it. Adapting the idea to a different situation reveals whether you understand the decision behind it.

Use resources that address the diagnosed gap. Internal service documentation may answer a production question better than another general frontend course. A foundational course can still be valuable when repeated difficulties reveal missing language or browser knowledge.

Make learning part of the work plan

Ask for a bounded investigation tied to a decision: “I need to establish whether this service can distinguish rejection from uncertain completion before we commit to the recovery design.” Give the output and explain the consequence of skipping it.

Agree on a review point and a feedback partner. Keep useful notes close to the code or decision so another person can find them later. If learning time is repeatedly displaced, discuss the capacity problem with your lead; a better personal study schedule cannot resolve an unfunded organizational responsibility by itself.

Use AI to expose and examine alternatives

AI assistance can help generate contrasting layouts, identify unmodeled states, explain unfamiliar code, or propose checks. Supply the decision and constraints. Ask it to identify assumptions rather than treating its output as established product knowledge.

An example request for this case is:

Review an assignment flow where the service may save a change before its response is lost. The client can read the latest record, but has no operation-status endpoint. Identify what the client can know, where another user’s changes could make reconciliation ambiguous, and which assumptions need service-owner confirmation.

The response may reveal useful questions. It does not confirm the actual contract, replace a service review, or create participant evidence. Likewise, generated personas and simulated interviews can help rehearse research questions, but they cannot validate a user need.

Use only tools and data approved for your work context. Provide sanitized examples where necessary, and review whether logs, screenshots, or source excerpts contain confidential information before including them.

Verify the artifact at the level of its consequence

Consider an AI-generated error handler that retries every failed assignment automatically. It may compile and pass a happy-path test. It still violates the uncertain-outcome reasoning from chapter 6.

Explain the proposed behavior in plain language, inspect the relevant code, and run a scenario capable of exposing the problem. Review generated interface content against known service outcomes. Check keyboard and assistive-technology behavior where the change affects them. Ask a specialist about guarantees outside your knowledge.

If you cannot explain or modify the generated implementation, reduce its scope until you can. The professional obligation follows the work you accept into the product.

Keep a small learning record

ItemExample
GapI treated a failed response as a failed write
PracticeImplement uncertain-outcome recovery in the assignment example
FeedbackA service engineer challenged the latest-record assumption for concurrent users
RevisionDescribe the limitation and propose an operation-status contract
Transfer checkApply the same reasoning to a payment or approval example without copying the first solution

Use the record to choose the next task. It does not need a score for every professional competency.

Practice: complete a learning cycle

Choose one difficulty exposed by the book’s case or your own project. Define a small task, a feedback partner, and observable readiness criteria. Complete it, revise it, and apply the idea in a second context.

If you use AI, retain a concise account of the useful contribution, an assumption you checked, and a correction you made. If you do not use AI, review a supplied or peer-written artifact with the same verification discipline.

Review your work

Show what changed in your ability, not only which resource you completed. Explain the new artifact independently and identify a limit of your knowledge. Your next practice task should follow from evidence of need.

The final chapter assembles the work into a professional account that another person can assess.

Found an error? Report it.