INNOVATION HUB · DESIGN THINKING
Human Skills and AI in Design Thinking
People first. AI on purpose.
Design Thinking has always been about people. You understand a problem by listening to the people who live with it. You solve it by working with others who see things differently than you do. That part of the work has not changed.
What has changed is AI. It can now draft, sort, summarize, and come up with ideas faster than any team. That makes the human side of design thinking matter more, not less. Someone still has to listen, notice what is really going on, disagree well, and decide what is true. AI cannot do that part for us.
This page is for students and instructors at the Innovation Hub. It walks through each stage of our design thinking process. At every stage, it highlights relevant human skills and how AI can help when used intentionally to augment what we already bring to the process. Instructors can use it to plan lessons. Student teams can use it as a guide throughout their design thinking project.
Working with people. Working with AI.
Human skills
How we work with people. In design thinking, people come first — both the people we design for and the people we design with. This is a key part of the work AI cannot do for us. That means naming what we and others are feeling, listening to understand, noticing when stress pulls us off course, being honest about our own part when conflict comes up, and staying present with a team when the work gets hard.
AI Fluency
How we work with AI in our classes, we think of AI as the last team member. It joins after the people, and the team decides together what it does. That means deciding when to use it, asking for what we need clearly, checking what it gives back, and being open about how we used it. Used this way, AI adds to what the team brings instead of replacing it.
The 4 Ds of AI Fluency
AI Fluency means using AI well. That covers getting good results, working efficiently, and using it ethically and safely. Sometimes AI does a task for us, sometimes it works alongside us, and sometimes it takes a series of steps on its own after we set it up. In every case, the same human skills decide how well it goes. Decision-making, teamwork, collaboration, emotional literacy, and self-regulation bring the context AI does not have. The people we are designing for, what they told us, and what our team has already learned. That context is what makes AI's output useful here, for this problem and these people.
For this, we use the 4D AI Fluency Framework. The four Ds are not steps you do once. Teams move through them again and again as a project goes on. The examples below follow a team whose project focuses on how first-year students keep track of deadlines. The 4D AI Fluency Framework was developed by Prof. Rick Dakan and Prof. Joseph Feller in partnership with Anthropic.
D1 · Delegation
Deciding whether AI should help with a task at all, and what to keep for ourselves. Some tasks go faster with AI. Others only work when people do them, because the value comes from doing the work ourselves. Delegation means making that choice on purpose before the work starts, and checking it again as the work goes on.
In design thinking. At the Empathize stage, the team could ask AI to describe how first-year students feel about deadlines. It would answer quickly, but it would be a guess, not something real students said. So the team interviews six students themselves. Later, at the Synthesize stage, they sort their notes into themes by hand first, because that is where they start to understand what they heard. Only then do they ask AI to check for anything they missed.
Ask yourself. If AI did this step for us, what would we stop noticing, learning, or understanding?
D2 · Description
Giving AI enough background to be useful. That means saying what the goal is, who it is for, what you already know, and what a good answer looks like. If the first answer misses, you explain what to change and try again.
In design thinking. At the Ideate stage, a weak prompt is "Give us ideas for a study app." A stronger prompt uses what the team learned in earlier stages. "We interviewed six first-year students. They told us they lose track of deadlines across different classes, and that they already have too many apps to keep up with. Give us ten ideas that do not add another app."
Ask yourself. Did we tell AI what we learned from the people we talked to?
Ask yourself. Did we tell AI what we learned from the people we talked to?
D3 · Discernment
Checking what AI gives back before we use it. AI can sound confident and still be wrong, leave people out, or repeat common assumptions. Discernment means reading AI's answers the way we would read a classmate's draft, carefully and with the evidence in front of us.
In design thinking. At the Synthesize stage, the team asks AI to look over its notes for themes it might have missed. AI suggests that students want more reminders on their phones. The team goes back to its notes and finds that two students said the opposite. They already have too many apps and notifications to keep up with. The team drops the suggestion and adds a theme that matches what the students actually said.
Ask yourself. Can we point to where someone actually said this?
D4 · Diligence
D4 · Diligence
Taking responsibility for how we use AI and being open about it. The team, not AI, is responsible for every claim in the final work. Diligence means keeping track of where AI helped and making sure nothing it made up gets presented as something people told us.
In design thinking. At the Pitch stage, AI can help design the slides. In one draft, it adds the line "Most first-year students miss at least one deadline a semester." None of the six students interviewed actually said that, and the team has no source for it, so they cut it. On the last slide, the team notes where AI helped. It checked the team's themes, suggested ideas, and polished the wording.
Ask yourself. Can we point to the person or source behind every claim in our final pitch?
Discernment and Diligence depend most on human skills. AI does not replace any stage of design thinking. It adds a new habit inside every stage. People do the work first, AI adds to it, and then people check what it gives back. AI can help point out what to check, but it cannot be the final judge. It was not in the room, and it does not answer for the work. When we check, we speak for the people we interviewed, making sure their voices are not replaced by a confident guess.
The ten design thinking stages, with people and AI
For each stage of the Innovation Hub design thinking process, here is the human skill that matters most, where AI Fluency comes in, and one idea for instructors.
1 · How Might We
Turn a broad topic into a question worth working on.
Human skill. Curiosity. Staying with the question before rushing to answers. This matters more with AI, because AI will answer whatever question it is given, even a weak one. Ask it "How might we build an app to help students track deadlines?" and it will hand back ten apps without asking whether students want another app. "How might we help first-year students keep track of deadlines across their classes?" leaves room to find out what they actually need. AI can produce answers. Deciding which question is worth asking is still the team's job.
AI Fluency. Delegation and Description. Each team member writes three How Might We questions on their own first. The team picks its strongest one and asks AI to rewrite it five ways: broader, narrower, from a student's point of view, from an instructor's point of view, and without assuming a solution. The team compares the versions and chooses. AI does not choose.
For instructors. Have each student write three questions before anyone opens AI. Then give AI one of the weaker questions and read its answers together. Ask, "What did AI assume because of how the question was worded?" Students see that AI answers the question it is given without questioning it. Questioning the question is a human skill, and it matters more as AI becomes part of the workflow.
2 · Team
Become a working team.
Human skill. Saying how you work best and what you need from others. For example, "I think best on my own first, so please send the agenda the night before," or "I go quiet when I disagree, so please ask me directly." This matters more with AI, because it is easy for each person to go off and work with AI alone. The team can end up with four separate projects instead of one. AI cannot tell you what your teammates need. Only they can, and only if someone asks.
AI Fluency. Delegation and Diligence. AI stays out of getting to know each other. Before the project starts, the team writes down how it will use AI. For example, "We do our own interviews. We write our own ideas before asking AI. We note every time AI helps."
For instructors. Have each person finish two sentences out loud. "When we use AI, I need the team to..." and "I will know we are leaning on AI too much when..." Teams add their answers to their team agreement. Check back mid-project and ask whether anyone has been working with AI alone instead of with the team.
3 · Plan
Map out the work and the deadlines.
Human skill. Being honest about time and workload. It is easier to say "I can do that" than "I have two exams that week." With AI, it is tempting to promise more, because AI makes the work feel faster. But AI does not speed up interviews, team meetings, or hard decisions. Honest plans make room for the human work that still takes time.
AI Fluency. Description and Discernment. The team gives AI its project deadlines and each person's busy weeks, and asks for a draft timeline. Then the team checks it. Does it leave enough time to set up interviews? Did it put heavy work on someone's exam week?
For instructors. Have teams ask AI for a project timeline, then build their own from each person's real schedule. Compare the two. Ask what AI's plan left out. Usually it is the human work, like setting up interviews, waiting for replies, and meeting as a team. Then ask each person to name one week they already know will be hard.
4 · Empathize
Talk with real people affected by the problem.
Human skill. Listening to understand, not to reply. When a student says “I just forget," the next question is "Tell me about the last time that happened," not "Have you tried a planner?" This matters more with AI, because AI can describe what users probably feel in seconds, and it will sound right. Only real listening reveals what is true for these students, including things no one would have guessed.
AI Fluency. Delegation. The team does not ask AI to pretend to be a first-year student or guess what students think. It interviews real students. Before the interviews, AI can suggest open-ended questions, and the team decides which ones to use. Deciding not to use AI is also a Delegation choice.
For instructors. Before interviews, have teams ask AI to describe how first-year students feel about deadlines, and save the answer. After the interviews, compare it with what students actually said. Ask what real listening found that AI either missed or just made up. The comparison shows students why the interviews are critical.
5 · Synthesize
Make sense of what you heard.
Human skill. Keeping your own assumptions apart from what people actually said. A team member may be sure students want more reminders because that is what they would want. This gets harder with AI because it often suggests the most common answer and says it with confidence. When AI and your own assumptions agree, it can feel like proof. The only proof is in your notes, in what your users actually told you.
AI Fluency. Delegation and Discernment. The team sorts its notes into themes by hand first. Then it asks AI to look for anything it missed. AI suggests that students want more reminders on their phones. The team checks its notes and finds two students said the opposite. They already have too many apps and notifications. The team keeps only what the evidence supports.
For instructors. Give teams one AI-suggested theme and ask them to prove or disprove it using only their interview notes. Every theme on the wall needs a quote next to it. If no one can point to a quote, remove it whether a person or AI suggested it.
6 · Reframe
Restate the problem based on what you learned.
Human skill. Letting go of the problem you walked in with. The team started with "How might we help first-year students keep track of deadlines?" What they heard was less about forgetting and more about having too much to keep up with. AI makes it easy to keep polishing the first problem, because it will improve any statement you give it. Changing direction takes a person willing to say, "We had this wrong."
AI Fluency. Description and Discernment. The team gives AI its themes and key quotes and asks for five new ways to state the problem. One version is "How might we help first-year students see all their deadlines in one place without adding another app?" The team picks the version that best matches what students said.
For instructors. Put each team's first How Might We question next to its reframed one. Ask what they heard that made them change it. Then ask whether AI helped them change direction, or mostly helped them polish the question they started with. Talk about why changing direction takes a person.
7 · Ideate
Come up with as many ideas as you can.
Human skill. Building on other people's ideas and staying open when your own idea gets passed over, even when AI's idea is the better one. "Yes, and..." keeps ideas moving. "That won't work" stops them. AI can produce more ideas than any team, so the team's value shifts. The strongest ideas often come from a person connecting an AI suggestion to something they heard in an interview. That only happens when teammates keep building on each other.
AI Fluency. Description. Each person writes down their ideas first. Then the team gives AI what it learned. "We interviewed six first-year students. They told us they lose track of deadlines across different classes, and that they already have too many apps to keep up with. Give us ten ideas that do not add another app." AI adds range. The team's ideas and AI's ideas go on the wall side by side.
For instructors. Have teams use one sticky note color for their own ideas and another for AI's. Then ask each person to build on one AI idea by connecting it to something they heard in an interview. Later, look at which ideas made the final choice. Were they people's, AI's, or a mix?
8 · Select
Choose the idea to move forward.
Human skill. Disagreeing openly and respectfully, then committing together. "I see why you like it, and I'm worried it won't work for students who commute" is honest and keeps the conversation going. This matters more with AI, because AI can make any option sound good. A well-written list of pros can end a discussion too early. Honest disagreement between people is what tests an idea before users do.
AI Fluency. Discernment. The team can ask AI to list pros and cons for its top three ideas. Then it checks each point against the interviews. AI might say an idea is easy to use, but only students can tell you that. The team makes the decision.
For instructors. Have teams ask AI for pros and cons of their top three ideas. Each person then names one point on AI's list they disagree with, and why. Before the team makes its choice, ask each person to say one concern out loud. A choice nobody pushed back on is not always one everyone agreed with.
9 · Build
Make a prototype and test it.
Human skill. Taking feedback without getting defensive. When a student says "I wouldn't use this," the useful response is "Tell me more," not an explanation of why they should. With AI, prototypes are faster to make, so teams can test more often and learn more. That only helps if the team is willing to hear that its idea is not working yet.
AI Fluency. Description and Discernment. AI can help the team create a quick first version, such as a sample weekly deadline sheet. The team tests it with two or three of the students it interviewed and changes it based on what they say, not on what AI suggests.
For instructors. Have teammates practice feedback in pairs. One plays a student who dislikes the prototype. The other practices answering with "Tell me more" instead of explaining. After real testing, ask what the team changed and whether those changes came from users or from AI suggestions.
10 · Pitch
Present your solution.
Human skill. Telling the story truthfully and staying steady under pressure. That includes being honest about what the team does not know yet. AI can make a pitch sound polished even when the evidence is thin, and audiences are learning to notice. What makes a pitch believable now is a team that can answer questions based on what it actually saw and heard.
AI Fluency. Diligence. AI can help tighten wording and slides. When a draft adds "Most first-year students miss at least one deadline a semester," the team cuts it because none of the six students said it, and there is no source. The last slide notes where AI helped. It checked the team's themes, suggested ideas, and polished the wording.
For instructors. Have teams swap pitches. Each team marks every claim in the other team's slides and asks, "Where did this come from?" Anything that traces back to AI instead of an interview, test, or source gets flagged. Then have teams practice answering hard questions without notes or AI, using only what they saw and heard.
We believe
People first, then AI’s input, then people again.
AI works best when people decide how to use it
Empathy happens between people
AI output is a starting point, not evidence
Being honest about AI use builds trust and transparency
Human skills and AI Fluency grow stronger together
Learn more
The 4D framework and its research background
Courses for students and educators from Anthropic Academy
Innovation by Design at the Innovation Hub
The 4D AI Fluency Framework (Delegation, Description, Discernment, Diligence) was developed by Prof. Rick Dakan (Ringling College of Art and Design) and Prof. Joseph Feller (University College Cork) in partnership with Anthropic and is used here under a Creative Commons BY-NC-SA 4.0 license.