Designing the future: A deep dive into generative AI and design thinking for innovation
Can generative AI tools like ChatGPT and Midjourney complement existing approaches like Design Thinking and Experimentation to help us innovate better?
AI is changing the business environment – are you ready to adapt?
“It is not the strongest of the species that survives, not the most intelligent that survives. It is the one that is the most adaptable to change.” Charles Darwin
In my business, we’ve been exploring and experimenting with generative AI tools like ChatGPT and Midjourney to see how they can complement existing approaches like Design Thinking and Experimentation to help us innovate better.
To date we’ve focused on Generative AI’s applicability to the front end of the innovation journey where customer needs are researched, insights are distilled, solutions are ideated, prototyped, and tested, and business models are shaped.
Our experiments
The first set of experiments we’ve run have been testing generative AI on real-world, but non-live projects. Noting, that it was critical we didn’t input any confidential client information into the AI machines.
These real-world experiments have been on industries and categories we have experience in, enabling us to confidently assess and rate the quality of the outputs.
What have we learned?
Here is what we’ve learned by stage, starting with customer discovery.
Discover
The first stage of your innovation journey after defining your project challenge and gathering existing research and knowledge on the category, trends, customers, brands, and competitors is to fill in the gaps you’ve identified through fresh customer research.
We found ChatGPT was a useful starting point for helping draft a research brief, research plan, and discussion guide, but it’s not in the same league as working with a specialist researcher, experienced marketer, or innovator.
It also provided some helpful suggestions for customer segments to interview – a useful stimulus to complement your and your teams’ thoughts.
Whilst ChatGPT was satisfactory at identifying the most common needs and pain points for a category, it is not a real human and lacks real-world experience beyond its training data, which limits the depth of customer understanding and observations it can provide.
Distil
In Distil we unpack, share, and synthesise our findings from the Discovery stage and then craft the synthesised observations into compelling customer insight statements to inspire and focus idea generation.
We found this to be one of ChatGPT’s weakest stages, which is logical as it is a very human stage that requires and utilises empathy, context, curiosity, inferring, interpretation, intuition, sensemaking and so on.
Its attempts at generating Empathy Maps resulted in very generalised statements that lacked real-world experiences, depth, and richness.
In terms of generating customer insight statements, the tool was reasonable at coming up with customer descriptions and needs but limited at getting to the deeper ‘why’ behind these needs and coming up with meaningful insights. We also found that many of the statements were verbose and used language and buzzwords that most humans wouldn’t use in an everyday conversation.
Based on our experiences ChatGPT would only be helpful in this stage as a basic cross-checking tool – have we missed any key needs or pain points, and for content and copywriting inspiration? Like offering alternative wording and helping to refine customer insight statements.
Ideate
In Ideate, we incubate on the customer needs and insights before then generating masses of ideas which are harvested down to a shortlist of the top ideas. These shortlisted ideas are then refined and fleshed out into idea canvases ready for briefing prototyping and testing.
We found this to be where ChatGPT could be most useful. Key uses included:
- Generating initial lists of ideas for kicking off both individual and group brainstorms.
- Generating stimulus for inspiring new ideas. ChatGPT could provide a wide range of examples for our ‘Best in World’ technique.
- Building and iterating on your ideas. It can highlight potential weaknesses in your ideas, and you can use it to iterate on your ideas with techniques like SCAMPER.
- Providing an alternative perspective on ranking ideas. However, I wouldn’t recommend using it for your final ranking and selection of ideas, as it lacks real-world experience and critical thinking skills.
- Fleshing out and writing up your top ideas. ChatGPT provided a satisfactory starting point, or cross-check, for fleshing out and writing up your top ideas into an Idea Canvas.
Prototype
In the prototype stage, we bring our ideas to life to make them testable in the quickest, cheapest, and most low-fidelity way required.
We explored the use of Midjourney for this stage. Midjourney is a generative artificial intelligence program that understands user prompts to create lifelike and realistic images.
We’ve found it to be useful for creating fast high-quality images for paper prototypes like concept boards and storyboards. It is also useful for creating persona and customer journey map visuals.
The biggest drawbacks with Midjourney are, firstly, everything you create is publicly viewable (unless you buy the enterprise license). Secondly, current laws indicate you can’t own the IP for AI-generated work.
Finally, whilst Midjourney enables just about anybody to generate images, engaging a skilled designer to use Midjourney will result in the production of this work, faster and better than a non-designer or novice. Skilled designers also offer much more than the ability to just rapidly draw a person or object, they are experts in visual communication as well as understanding the intricacies of branding, designing experiences, and so on.
Test
In the test stage, we want to find out if our ideas (prototypes) are firstly desirable, and then once we’ve validated this whether they are also feasible and viable. Testing is usually an iterative process of testing your prototypes with customers (and other users and stakeholders), getting feedback, making modifications, and going through the loop again.
Hopefully, I’m stating the obvious here — ‘testing’ your ideas on a generative AI platform like ChatGPT is not testing them with actual customers, nor are you currently able to upload actual prototypes for feedback.
However, what you can do is ask ChatGPT ‘What feedback would customer x have on these ideas?’ We found this useful as an initial stress test of our ideas and to provide some input for refining our ideas before building the prototypes and putting them into real testing with humans. We also found that the outputs from the ‘testing’ with AI customers provided some areas to possibly explore and probe with customers in the actual testing.
Business Model Design
In business model design, we want to flesh out our idea into an end-to-end solution covering the front, middle and back office to see if we have a desirable, feasible, and viable business model.
We found ChatGPT created a satisfactory starting point, or cross-check, for creating and refining a Business Model Canvas. If using it as a starting point, you’ll want to modify and tailor it based on your specific market and organisational considerations and your and your teams’ expertise and experience.
Conclusion
In conclusion, from our experiments to date, we’ve found that generative AI is complementary to how teams currently innovate and has the potential to enhance their innovation efforts to a moderate extent.
I see its main use as being able to rapidly offer extra inputs, stimulus, and content for the various stages of the innovation journey either as start points or cross-checks.
The 24/7 assistance is also an attractive feature in an age of WFH, the solopreneur, side hustles, portfolio careers and the like. Unlike most human collaborators, generative AI is available 24/7.
Whilst, it does have the potential to help teams innovate at lightning speed (one could sit at their desk and do a fully AI-generated Customer Discovery to Business Model project in a matter of hours rather than days or weeks) I would offer a word of caution here. Running a wholly AI project would result in cutting out the more time-intensive yet critical human skills like intuition, real-world experience, critical thinking, in-depth analysis, sensemaking, and collaboration; and human activities like cross-functional team sessions, workshops, customer research and testing, leadership showcases, retros and reviews to name but a few which are all crucial to innovation success.
A few other words of caution. generative AI does not have real-world experience or practical knowledge beyond its training data. It may unintentionally generate biased or non-neutral content, which could impact the objectivity of generated content.
And then there is the issue of confidentially and security. Discussing sensitive or proprietary information with generative AI may pose security risks as the models don’t have the ability to keep information confidential.
In terms of generating frameworks, processes, and ways of working it is not in the same league as most organisations’ internal toolkits, good innovation books, or experts such as specialist researchers, facilitators, designers, marketers, and innovators who can also offer much more.
What’s next for us?
Now that we’ve tested these tools out in a lab environment, we’ve started our next wave of experiments, which is safely and carefully using generative AI on actual projects and training workshops with clients. If you’d like to be involved in these trials, start using generative AI on your innovation projects, or have your team trained in their use, please reach out.
Nathan Baird is the founder of Methodry and the author of Innovator’s Playbook: How to create great products, services, and experiences that your customers will love!
