AI & UX: from strategy to execution

A three-part series on successfully implementing artificial intelligence in user experiences.

These three articles form a practical guide to designing AI experiences that truly add value. From strategic scoping to practical implementation and critical evaluation, each article builds on the previous one and offers concrete guidance for organizations that want to use AI without falling into familiar pitfalls.

Scoping AI: asking the right questions before you start

We are in the middle of the AI revolution, and every organization feels the pressure to keep up. ChatGPT-like chatbots are popping up everywhere, and new AI features are being pushed into many applications.

The result? Frustrated users, disappointed stakeholders, and AI implementations that create more problems than they solve. The hard truth is that 70% of AI projects fail, not because of poor technology, but because of poor scoping.

The “AI just because we can” trap

“Can we add AI to this as well?” It’s a question I hear often in the hallway. It’s the modern version of “can we make an app for this?” The problem lies in the mindset: technology looking for a problem, instead of the other way around.

Real AI success starts with a fundamentally different question:

“Which user problem can we solve better with AI than with UX alone?”

Netflix uses machine learning (a subfield of AI) because it’s impossible to manually curate relevant content for 260 million users. Google Translate exists because there is a real need for instant translation. These tools genuinely add value they make something possible that would otherwise be unachievable.

The value check: five critical questions

Before investing in AI, make sure you can answer the following:

1. What is the underlying user problem?

Not the symptom, but the real pain point. “Users can’t find what they’re looking for” is a symptom. “Users lack sufficient context to make good decisions” is an underlying problem that AI may help address.

2. Why is a UX-only solution insufficient?

If a better search function or improved information architecture solves the problem, why introduce the complexity of AI? AI is powerful, but also fragile and expensive to maintain.

3. Do we have the data to make AI successful?

AI without quality data is like a car without fuel. “We’ll collect data as we go” may sound smart, but in practice it’s little more than wishful thinking.

4. Can users understand and interact with the output?

An AI tool that generates perfect output but is impossible to interpret is useless. Users need to be able to understand, trust, and intuitively work with the results.

5. What happens when things go wrong?

AI fails. Always. It’s not a matter of if, but when. That’s why it’s crucial to understand the impact of errors upfront.

Some mistakes are relatively harmless: a poor Netflix recommendation or an irrelevant search suggestion might only cause frustration or wasted time. Others can be far more serious: incorrect medical advice, financial miscalculations, or flawed legal suggestions can cause real harm or even be life-threatening.`

When scoping AI, you must critically assess the risk level of potential errors and decide which safeguards, checks, or human intervention are required. This prevents you from introducing technology that seems helpful at first glance but ultimately creates more problems than it solves.

Balancing control and automation

One of the hardest aspects of AI scoping is finding the right balance between user control and automation. Fully automated systems are efficient but can make users feel they’ve lost control. Too much manual intervention, on the other hand, eliminates the benefits of AI.

The sweet spot lies in augmented intelligence: AI that enhances human capabilities rather than replacing them. Spotify’s Discover Weekly is a good example. AI suggests new music, but users decide what they save, skip, or replay. Control stays with the user, while AI enables discovery.

Transparency: the trust formula

Users only accept AI if they understand what it does and why. This doesn’t mean exposing algorithms, but it does require clear communication:

• What does the AI do? (“This tool analyzes your spending patterns.”)
• Why does it do this? (“To help you save money.”)
• How reliable is it? (“Based on 1 million similar profiles.”)
• What can you do with it? (“You can accept, adjust, or ignore the recommendations.”)

No hype, just strategy

AI scoping is ultimately about asking the right questions. It means saying no to shiny AI features that don’t solve real problems. It means choosing boring but effective implementations over sexy but useless functionality.

Organizations that succeed with AI are not the ones that adopt it first. They are the ones that work most thoughtfully: focusing on user value and treating AI as a means, not an end.

The question isn’t whether your organization needs AI. The question is where, when, and—most importantly—why. Ask the right questions, and the rest will follow.

Curious whether your organization is ready for AI? Take our AI Scan and quickly discover where you stand.