Think first, AI second: Using AI in Design Thinking
Image by ChatGPT

Think first, AI second: Using AI in Design Thinking

Ian Smith
Ian Smith

I've been thinking a lot recently about the role of AI in creativity — particularly in the context of structured approaches like Design Thinking. There's a real tension here that continues to evolve as AI capabilities continue to advance at speed. This article represents my current thinking as of early 2025. However, I hope that you will find the fundamental principles outlined here to be useful into the future, as AI becomes ever more integrated into our work and lives.

Used naively, AI can crowd out your own thinking and potentially narrow the scope of your ideas, especially if some of its ideas seem good right out of the gate. As humans, we can easily get fixated on these and feel as though the question is answered without ever having fully engaged with the problem ourselves. Used well, however, AI can help you see more clearly — and more broadly — than you otherwise would.

The Design Thinking Loop

When I talk about Design Thinking, I am really talking about IBM's Enterprise Design Thinking method, which is what I used when I first learned Design Thinking. I find that this approach recognises, better than other frameworks, the iterative nature of both designing, and building. In their method, IBM talk about the Loop:

The Design Thinking Loop

This shows the stages of the Design Thinking process as an iterative loop, moving from observation, to reflection, to making, then back through further reflection, to observation of how users interact with what you made, and so on back around.

Observe

The Observe stage is all about talking to the people you're trying to help, getting into their shoes, and observing how they currently solve the problems you're addressing. This is crucial to making sure that you're solving their actual problems, and not simply working from the team's preconceptions. Later, you will be observing how they interact with your prototype solutions, and using this knowledge to refine these, or even discard them and start again.

AI can be a powerful assistant during the Observe stage, particularly for summarizing and synthesizing data from user interviews, field observations, and other research. When you have multiple lengthy transcripts or observation notes, AI can help identify patterns, recurring pain points, and common user behaviors that might otherwise be missed. This synthesis capability allows your team to work with larger datasets and potentially uncover insights that might be buried in the volume of information.

That said, it's important to remember that AI's understanding is only as good as the data you provide it. If you start to rely on AI having direct knowledge of your users and their problems, then you run a big risk of being misled. If the problem your team is addressing coincides with what the AI has been trained on, it may well have some useful insights to contribute. However, this is no substitute for personal knowledge, as it's most likely that you're addressing a specific version of the problem in a specific context for specific users.

Current AI technology has limitations in making contextual connections. Unlike humans, AI cannot remember specific facts that become relevant to your project later on, especially facts that seemed unimportant when you first encountered them. Many of these insights might not even be documented anywhere. The human brain excels at making these unexpected connections between seemingly unrelated information, which is why it's important for you to engage directly with the people you're trying to help, and the problems they face, taking ownership of the research and ideation process rather than delegating it entirely to AI.

Make

The Make stage is all about prototyping, and AI can very often help here. For example, I often use AI to help rapidly prototype an idea with code, and as AI is integrated into more and more business processes this can expand into real world prototyping.

If you have a well formed and clear idea, AI can make this process much faster, allowing you to quickly put high fidelity prototypes into the hands of your users and accelerate the learning process.

Reflect

However, in this article I'm focusing on the Reflect stage, in which the team comes together to make sense of the insights and outputs from the other stages. It's unlikely that every team member will have the opportunity talk to every user, so this step allows the team to document a shared view that they can all work from.

This step is where we find the Design Thinking workshop that everybody loves to participate in, but the workshop can seriously fail the team, if it is performed in isolation without the other stages.

In the workshop we focus on activities that help us document and make sense of as a team what we have learned. These include workshop activities like Empathy Mapping and As-Is Scenarios, as well as Big Ideas Vignettes and Prioritisation Grid to name but a few.

AI can be really helpful at this stage, but it can also limit the team's thinking and narrow the scope of potential innovation. So how can AI help us with Reflect, and what should we watch out for?

Principles

It's tricky to balance using AI's capabilities while maintaining awareness of its limitations. This may seem contradictory, but this balance actually forms the foundation of successfully integrating AI into any process or domain. It's similar to other important balances in Design Thinking, such as between divergent and convergent thinking (exploring many ideas vs. narrowing to solutions) or between innovation and feasibility (what's desirable vs. what's possible). These create natural tensions that we need to recognize and manage.

Here are a few principles, that I think will help.

Think First, AI second

If there's only one thing that you take away from this article, I hope it will be this. In fact, it is so important to me, that I even put it in the title! The message here is to perform each activity first without AI. Make sure your thoughts and ideas and those of the team are all out there before you to go to AI for more, so that your uninfluenced thinking is preserved.

Once you're done, give the AI any inputs that are already available, such as screenshots from MURAL, or a photo of the sticky notes from previous activities and ask for its help. If there are good ideas or new thoughts, add them - but ensure that you mark them as AI generated.

Ask for specific contributions, not general solutions

You should frame AI prompts to target gaps or expand specific areas rather than asking for complete solutions. Prompts beginning with phrases like "What might we be missing about..." or "Generate additional perspectives on..." can help fill gaps in the team's thinking - pulling out ideas or thoughts where team members had perhaps discarded, assuming that everybody would know.

Follow the usual AI prompting advice, and avoid requests that are vague or too general, such as "help us with our empathy map". Take the results of prior steps, or user interview notes, or any other inputs that were used by the team in the current activity, and give these same inputs to the AI.

An example of a prompt might be:

We are working on an empathy map for <insert persona details here>. Attached is our current empathy map, user interview transcripts, and persona documentation. Please:

  1. Review our empathy map and identify 2-3 potential gaps in each quadrant (what they say, do, think, feel)
  2. Generate 5-7 specific questions that would help us uncover missing insights, especially around <specific area of interest>
  3. Highlight which 2-3 questions you believe would yield the most valuable missing information
  4. Format each question as an open-ended discussion prompt that a facilitator could pose to the team

Our workshop has 15 minutes remaining for this activity, so please prioritise depth over breadth.

I like getting the AI to provide the output as questions, because these can be used as building blocks for team discussion, rather than as final answers.

Use AI for synthesis, not ideation

At the current stage of AI development, human teams that aren't using AI often bring strengths in divergent thinking and creating novel connections, particularly when addressing domain-specific challenges. That's not to say that AI has no value here, and the value is generally increasing over time, but you, as a human, still bring contextual understanding and lived experience that can give you an edge in creative ideation. Remember that the AI is processing a large dataset of text written by humans, but you have actual knowledge, experience, emotions, and a perspective that an AI can never have because it is not alive and has no embodiment or senses or experience of life.

So work with the team to generate the initial ideas, then use AI to organise, categorise, and highlight themes. Ask your AI to summarise your team's outputs for other stakeholders. This can provide great benefit, freeing up team members for tasks that only humans can do.

AI in Specific Reflect Activities

Here I have created some bullet points for a few of the more common Design Thinking activities, to help you and your team get the most out of AI and avoid pitfalls.

Empathy Mapping

Empathy Map

The Empathy Map is a core artefact that is often revisited and updated throughout a project.

  • First approach: Complete the empathy map exercise as a team without AI.
  • AI contribution: Have AI analyse the completed map to identify missing perspectives or contradictions, and frame useful new questions to ask the people we're looking to help. If you have the right input data, AI might help you notice differences between what people say they do, vs what they actually do.
  • Pitfall to avoid: Don't ask AI to generate an entire empathy map from scratch whether based on minimal data, or even more complete data such as user interviews, etc. Work on it as a team first.
  • Success pattern: Use AI to challenge assumptions in your map by asking "What if our user actually felt..."

As-Is Scenario Mapping

As-Is Scenario Map

As-is scenario mapping is often the most demanding part of a Design Thinking process, as scenarios are almost never simple.

  • Begin by: Mapping the current user journey based on observation data and team knowledge.
  • AI contribution: Identify bottlenecks or pain points that might not be immediately obvious. Also, AI can help spot unstated assumptions in your scenario map, although this demands much of the AI, and it's views are more likely to be wrong.
  • Pitfall to avoid: When you're engaged in the hard work of scenario mapping, it can be very tempting to resort to AI early. Don't fall for this, as you are robbing yourself and the team of the new understanding of the problem that is gained by going through the process.
  • Workshop technique: After completing the map, ask AI to take the user's perspective and identify what's most frustrating.
  • Practical tip: Use AI to suggest potential measurements for each step to quantify friction points.
  • Key benefit: AI can help spot patterns across multiple scenarios that suggest common underlying issues.

Big Ideas Vignettes

Big Ideas Vignettes

This is the fun part. Don't let the AI take away your fun.

  • First step: Have the team brainstorm big ideas independently before any AI involvement.
  • AI enhancement: Use AI to add new ideas, and expand promising ideas with additional features or implementation approaches.
  • Pitfall to avoid: Don't let AI's ideas have undue influence over which ideas the team pursues. AI speaks fluently and confidently, but it's important to see through the words to the idea beneath and compare ideas, rather than the fluent way in which they're expressed.
  • Workshop integration: Generate 2-3 AI variations of each team-created big idea, then discuss differences.
  • Key principle: Use AI to diversify thinking, not narrow it.

Prioritisation Grid

Prioritisation Grid

When prioritising, we examine feasibility of the idea, and impact on our user. AI will have views on both of these, but again it's important to establish a baseline as a team first.

  • Begin with: Complete initial prioritisation as a team to establish baseline consensus.
  • AI application: Analyse prioritisation choices for potential biases or inconsistencies.
  • Effective technique: Have AI try to promote and defend lower-ranked ideas to ensure thorough consideration.
  • Important note: Final prioritisation decisions should always remain with the human team.

Balancing Human and AI Contributions

Creating a Transparent Process

When integrating AI into your Design Thinking workshops, transparency is essential. Document clearly which ideas originated from team members versus AI suggestions, making this distinction visible in your workshop outputs. Before beginning, co-create clear guidelines with your team about when and how AI input will be sought during each activity, and write them down. This creates appropriate expectations and prevents over-reliance.

Consider appointing an "AI sceptic" role within your team—someone responsible for thoughtfully challenging AI-generated contributions and ensuring they truly add value rather than simply being accepted due to their polished presentation. This role can rotate among team members, encouraging everyone to develop a healthy critical perspective.

After each activity, create dedicated space for the team to reflect on how AI influenced their thinking. Ask questions like: "Did the AI suggestions expand or narrow our thinking?" and "Were we too quick to adopt AI ideas over developing our own?" These reflections will help refine your approach over time.

When to Avoid AI Input

While AI can be valuable, there are specific situations when it's better to rely solely on human input. When your team is struggling to generate initial ideas, resist the temptation to immediately turn to AI. This moment of struggle is where human creativity can be most inspired, and often leads to breakthrough thinking. Short-circuiting this with AI can prevent the team from developing their own creative muscles.

Some challenges involve more sensitive user issues where empathy and lived experience are crucial. AI lacks the emotional intelligence and contextual understanding that humans bring. In these cases, greatly increase your emphasis on the primacy of human perspectives, using AI only for supplementary analysis.

When workshop time is extremely limited and team alignment is the primary goal, introducing AI contributions can sometimes create more discussion than resolution. In these situations, focus on using AI to pull together the human team's thinking rather than on using it to introduce new elements.

Finally, if you notice signs that your team is becoming overly dependent on AI suggestions—such as deferring to AI before attempting their own thinking or attributing greater authority to AI ideas—it may be right to temporarily limit AI usage and refocus on human-centred ideation.

AI's Potential

This article has focused primarily on integrating AI into existing Design Thinking processes, but it's worth considering the ways in which AI, as it advances in capability, may be able help us develop entirely new approaches to human-centered design.

Apps such as Mural are already integrating AI into their toolsets, enabling automatic clustering of sticky notes, and even suggesting new sticky notes based on the content of the existing notes. It seems only a matter of time before AI makes the leap into becoming a full team member, able to contribute in each stage of the Design Thinking process.

Nevertheless, regardless of how AI transforms our methods, the core principle of designing for human needs and experiences remains constant. The tools and techniques may evolve, but the purpose of Design Thinking endures.

Conclusion

AI can significantly enhance the Reflect stage of Design Thinking by expanding perspectives, identifying blind spots, and helping teams process complex information. The key to success lies in using AI as a complement to human thinking, not as a replacement. By starting with clear human contributions before bringing in AI assistance, you preserve the essential human-centred nature of Design Thinking.

Remember that Design Thinking is fundamentally about human-centred problem solving, and while AI can analyse patterns and generate suggestions, the deep understanding of human needs remains the domain of the human design team. By combining human creativity and empathy with AI's analytical strengths, we can achieve our desired outcomes much more quickly, and possibly even with higher quality.

As AI capabilities continue to evolve, so too will its role in Design Thinking. By establishing thoughtful practices now for integrating AI while preserving human agency and creativity, design teams can harness these powerful tools without losing the essence of what makes Design Thinking valuable in the first place: its unwavering focus on the humans we're designing for.

If you'd like to talk to me about using AI in your Design Thinking practice, or otherwise implementing AI in your organisation, please get in touch using the button below.