Not a day passes by without an impressive AI demo making its rounds across Slack channels and LinkedIn posts. They highlight just how incredible the capabilities of AI systems can be. But rarely do these demos turn into actual viable products, often disappearing into the void weeks after the release.
On the other hand, while many teams are spending a lot of time debating and discussing AI, only few teams actually integrate AI into their products. And the issue isn't the lack of design skills or implementation hurdles: it’s the product-market-fit gap that is often overlooked or dismissed in favor of a super-intelligent text box.
Designers envision AI systems that often can’t be built, or can’t be reliable enough to take the human out of the loop. AI must prepare, ask a human to review or confirm, then accept or seek alternatives. And autonomous AI agents today are very fragile and very slow yet, often poorly imitating human behavior.
AI has a remarkably high sustainability cost. Not every product needs AI sparkles. Often, they are swooped in with a big new company-wide initiative, just to be entirely ignored by users upon launch. And once AI isn’t reliable enough, typing "speak to human" is their only chance to make any sense from that super-intelligent text box.
🤔 Designers often envision AI systems that can’t be built.
🚫 Most ideas require near-perfect accuracy to be useful.
🤔 Data scientists envision AI systems that have little interest.
🚫 Most products fail fast as they solve a wrong problem.
✅ Useful AI products are grounded in real-life applications.
🚫 AI is great at first drafts, not good enough as final output.
✅ AI has at best around 85–90% accuracy in its responses.
✅ Hallucinations and forgetfulness are difficult to resolve.
✅ AI still “forgets” context from the middle of a conversation.
🚫 Poor data → poor outcomes: data cleansing is critical.
🤔 We ran out of human data to train LLMs → synthetic training.
🤔 Around 95% of all content will be AI-generated by 2027.
✅ AI systems are becoming reasoning engines (agentic AI).
✅ Sparkles are out: AI slowly becomes an invisible part of UIs.
✅ Text boxes are replaced by UI controls, knobs, insight panels.
🚫 We can no longer distinguish human output vs. AI output.
🚀 Fantastic AI Examples:
Scispace (search + AI):
Dream Machine (inline action):
v7 Labs (AI auto-fill):
Exa (embeddings search):
DeepL (translation):
Elicit (research tables):
Miro (AI presets):
NotebookLM (scoping):
Claude (shareable URLs):
Perplexity (search + AI):
🤖 How To Design Better AI Experiences. With practical guidelines on how to add AI when it can help users, and avoid it when it doesn’t ↓
Many articles discuss AI capabilities, yet most of the time the issue is that these capabilities either feel like a patch for a broken experience, or they don't meet user needs at all. Good AI experiences start like every good digital product by understanding user needs first.
🚫 AI isn’t helpful if it doesn’t match existing user needs.
🤔 AI chatbots are slow, often expose underlying UX debt.
✅ First, we revisit key user journeys for key user segments.
✅ We examine slowdowns, pain points, repetition, errors.
✅ We track accuracy, failure rates, frustrations, drop-offs.
✅ We also study critical success moments that users rely on.
✅ Next, we ideate how AI features can support these needs.
↳ e.g. Estimate, Compare, Discover, Identify, Generate, Act.
✅ Bring data scientists, engineers, PMs to review/prioritize.