A practical guide to combining generative AI with human-centered design

Toolset

Toolsets are the instruments and resources you use day-to-day, such as digital whiteboarding tools, design software, and facilitation tools.

In the context of AI, understanding the basics of how AI and LLMs work is crucial to maximising how you use the tool. LLMs work through three core concepts:

• Pre-trained data: LLMs are trained on large datasets from the internet, scholarly articles, and books. Understanding how an LLM has been created including its pre-trained data helps you recognise and manage for risks such as bias, as well as maximise the tool’s output.
• Prompt-output relationship: An LLM generates responses based on statistical relationships in your prompt. It predicts the next most likely word from its training data. This reliance on statistical relationships when paired with poor prompting is in part what can lead to outputs feeling two-dimensional or shallow, as it can flatten the human nuance of design research.
• Tokenisation: An LLM breaks down your prompt into what are called tokens. Misalignment can occur during this breaking-down process, so rephrase prompts if needed.

Considerations for designers:

• Model comparison: Know the strengths and weaknesses of the GenAI model or tool you are using.
• Ethical alignment: Ensure AI-generated content aligns with your ethical standards and be mindful of biases.

Mindset

Mindset is arguably the most significant attribute of a designer. It's what sets you apart from other roles within a multi-disciplinary team. A designer's mindset involves seeing things differently and approaching problems with a unique perspective.

Key mindsets for integrating GenAI into design practice

• Get curious and experiment: Embrace curiosity to explore new approaches with GenAI. For example, if your research plan usually starts with empathy interviews and affinity maps, see how GenAI can help diversify your methods and get you out of your comfort zone.
• Plan ahead: Map your traditional design workflow and identify where GenAI can be integrated. This is useful for activities like creating a service blueprint.
• Start rough, test early: Test initial ideas from rough or unfinished notes with GenAI to see what you get back, allowing for iterative improvements. This is very complementary to the above approach to ensure your new design workflow works the way you intend.
• Ask "what if" and "what’s next": Use follow-up questions to build and extend outputs further. For example, if you have some finished personas, consider how you might use AI to connect these personas to a stakeholder map, or use the personas to expedite the creation of a value proposition canvas.
• Iterate continuously: Don’t settle for the first result. Refine your prompts to improve the analysis and outcomes.
• Continuous learning: Use GenAI as a coach to learn complementary skills. For instance, enhancing your ability to articulate insights from data analytics can enrich your HCD practice. Embrace GenAI to expand your skillset and stay adaptable.

Skillset

Skillsets determine how well a designer executes their work by how you apply tools to design problems. For example, knowing when to use a service blueprint instead of a customer journey map.

In the context of GenAI chatbots, the key skillset is effective prompting. Here are some tips:

• Use AI personas: Start your prompt with an AI persona to guide the tone and structure for the output. For example, "Assume the persona of a communications professional who favours concise communication..." would be a useful persona for writing up a project summary. 
• Provide or ask for definitions: Ensure you and the chatbot are on the same page about frameworks.
  • For example, if you want to use the ‘Jobs to be Done’ framework, your first prompt might be “Assume the persona of a human centred design professional, tell me about the Jobs to be Done framework” and if you’re satisfied, your follow-up prompt will be something like “Apply the Jobs to be Done framework explained above to…”
  • If you aren’t satisfied with the explanation, you will need to explain it and confirm the understanding prior to proceeding into the exercise.
• Ask for reasoning or explanations: Treat the chatbot like a team member:  ask for rationale to deepen understanding and check logic. For example, "For each point, explain your reasoning and provide examples from the source material."
• Reduce hallucinations: If relevant, include instructions like "If the information isn’t available within the source material, mark as 'information not available.'" to avoid the AI tool inserting material that hasn’t been provided.

Ways to use GenAI in your day

This is by no means a comprehensive guide – these are just some of the ways that this service designer has used GenAI over the past year. Whilst these use cases can offer helpful inspiration, it’s important to highlight that the technology is advancing at a rapid rate so they will likely look very different a year from now.

Basic time savers that are transformative:

You can start with something as simple as…

• Use image-to-text to digitise hand-written notes from in-person workshops
• Use image-to-text to save typing up important information from diagrams, and leverage this text within larger pieces of research and synthesis
• Generating useful elements such as instructions, briefs, or user testing email invites.

It's important to use this text and the time you’ve saved to delve deeper. Reflect on those mindsets, and now that you have hand-written notes digitised, ask yourself “what if?” and “what’s next?”

AI from the start

Before starting anything whether it’s creating a workshop or project – dedicate a few minutes to chatting with AI to explore your approach, planning, and see if there is anything you haven’t thought of, you’ll be surprised how it helps to bring you to a more well-rounded activity or plan. The use of AI personas in your prompting can be valuable to get the most out of this approach.

Augmenting design processes

This service designer has had success in generating first drafts of customer journey maps and service blueprints, rapidly speeding up this process. The key things you need to do this successfully is to explain the swimlanes you need and then provide it with access to the complete collection of raw notes that you’ve gathered as part of your design research. As a first draft it can be very effective, however, you’ll need to make sure you check for hallucinations. It’s recommended as the designer (not the AI), to fill in the sentiment at each stage, as well as adjust the pain points or opportunities to preserve some human nuance.

Storytelling

This is where GenAI chatbots can really excel, and one example where this has been used very effectively was to generate a storyboard narrative for a complex multi-week process, where we wanted to communicate what the future experience would be like. The chatbot ingested rough notes on the future state process and was able to easily create a digestible storyboard split into 8 chapters, which helped to simplify what the process meant for them. It was also able to create personalised storyboards for each role within the process – meaning whether you’re customer service or the customer, you understood exactly what the future state could mean for you.

Continual and rapid adaptation

AI technology has developed rapidly since ChatGPT made headlines two years ago. With advancements like enhanced reasoning models and integration into various AI Agents, AI is no longer a stand-alone product but is becoming a part of our favourite design and non-design programs.

To succeed as a designer in this rapidly evolving landscape, it is crucial to adapt your toolset, skillsets and mindsets. You can always reflect and continue to evolve by asking yourself these questions:

Toolset:

Ask yourself…

• How can I understand data/technology underpinning the program I’m using?
• Why I’m getting the result or output I’m receiving with GenAI?
• Am I utilising the full capability of what is available to me?

Mindset:

• Nurture curiosity, be experimental, and take an iterative approach to learning.

Skillset:

• Identify gaps across your design skills. Seek feedback from peers, mentors or external research to validate your self-assessment – then explore how AI can help you to upskill with this.
• Identify complimentary skills you can use AI to help you upskill in.
• Through either external research or internal networks, find out how designers are using AI at your organisation.