NN/g’s AI Design Study Guide: What to Read First

Most AI design advice starts at the wrong end. It assumes you’re building the feature and jumps straight to prompt patterns and chat layouts. NN/g’s new designing AI products and features study guide does the opposite — its first section is about deciding whether AI belongs in the product at all.

The guide, published September 18, is a curated index of NN/g’s own articles and videos on designing AI products and features. It’s a reading list, not a tutorial. That’s exactly why it’s useful: it tells you the order the research house thinks you should think in, and that order is more opinionated than a link dump usually is.

It opens with a warning, not a technique

The framing line is blunt: adding AI doesn’t magically create value, and in some cases it’s a large investment for a small return. NN/g points back to Jakob Nielsen’s 2021 warning about overinvesting in new technology before understanding how it solves a real user need.

The hammer metaphor in the guide is worth sitting with. There’s nothing wrong with AI as a tool in your kit — the problem is hunting for a nail to hit with it. If you’re a product designer being handed an AI feature request, that reframing is the most useful thing in the whole collection.

Lead with the value the AI implementation offers, not with the AI itself. That’s the guide’s stated position, and it sets up everything after it.

The four things AI is actually good at

NN/g names four “superpowers” where AI improves products: content creation, summarization, basic data analysis, and perspective taking. That’s a short list, and its shortness is the point.

If your feature idea doesn’t land in one of those four buckets, the guide’s implicit question is why you’re building it. It also flags two patterns that rarely pay off: AI built for novelty, and chat interfaces adopted by default rather than because users need a conversation.

Narrowly scoped AI features, the guide says, are easier to understand and get better adoption. That’s a design constraint you can act on this week — cut the scope until the feature does one of the four things.

Where the prompting research lives

A large chunk of the collection deals with the articulation barrier: users can’t always say what they want, and they don’t know what the model can do. NN/g’s answer is hybrid interfaces — prompt input combined with graphical controls.

Three specific pieces are worth reading in sequence:

• Overcoming the Articulation Barrier in Generative AI Using Hybrid Interfaces — why prompt-plus-GUI beats prompt-only for image generation.
• Prompt Controls in GenAI Chatbots — the four jobs controls do: discoverability, education, constraint-setting, and follow-ups.
• Use-case prompt suggestions — suggestions have to be contextually relevant, personalized, and matched to the user’s experience level.

There’s also a piece on generative UI producing simple elements like buttons and checkboxes inside a chat, which reduces typing and memory load. If you’re sketching an AI feature right now, that’s the most directly applicable idea in the set.

The output-quality half nobody reads

The guide’s second half covers writing and presentation, and it’s the part most teams skip. NN/g’s position is that generative output still has to follow web-writing rules: concise, scannable, inverted pyramid, plain language.

Two findings here are worth quoting to a stakeholder. First, AI summaries of product reviews land well when they’re specific and transparent. Second, Amazon’s Rufus shows that even a valuable AI feature has low impact if users don’t notice it exists.

There’s also a short piece on the sparkles icon and why it’s ambiguous — a small thing that shows up in almost every AI feature shipped this year.

What the guide doesn’t give you

It’s a collection of links, so it inherits the gaps of the underlying articles. There’s no cost model, no build-versus-buy comparison, and nothing on evaluating a model vendor. If your decision is commercial rather than interactional, this won’t settle it.

It’s also NN/g’s own material only. That’s a feature if you want a consistent point of view and a limitation if you want the field’s disagreements.

How to use it as a product designer

Read it in three passes rather than top to bottom.

1. Before you commit to the feature: the value framing and the four superpowers. Decide whether your idea fits one of them.
2. While you’re designing the input: the articulation-barrier and prompt-control pieces. These change your wireframes.
3. Before you ship: the writing and discoverability material. This is where features quietly fail.

If you’re a freelancer scoping an AI feature for a client, pass one is your scoping conversation. Being able to say “this doesn’t map to any of the four things AI does well” is a cheaper argument than building a prototype nobody adopts.

Where it sits next to other learning paths

NN/g’s guide is free and primary-source, which makes it a different kind of resource from a paid course. If you want a structured credential rather than a reading list, Coursera carries ACE credit-recommended Professional Certificates, and Coursera‘s own September 16 post reports more than 50 of them and nearly $60 million in modeled potential tuition savings. That’s a different purchase decision — credit toward a degree rather than a design method.

For a portfolio, Adobe Portfolio is the low-friction place to show the AI feature work once it ships. Neither replaces the guide’s actual job, which is telling you when not to build.

The takeaway

The guide’s value isn’t the links — it’s the sequence. Decide whether AI adds value, scope it to one of four superpowers, solve the articulation barrier in the interface, then make the output readable and findable. Skip step one and the other three won’t save the feature.

Read the value section before your next AI kickoff. It’s the cheapest hour you’ll spend on the project.

Sources

• Designing AI Products and Features: Study Guide — NN/g, September 18, 2026
• From skills to degrees: New ACE data — Coursera blog, September 16, 2026