A Novel Counterpart: Using AI as Part of the Innovation Process


AI can be a powerful counterpart for your innovation process. That is only if we have ways to recognise and deal with its limitations and biases. At ThinkPlace, we’re experimenting with the use of AI in the research and ideation stage of design by using it to enhance our capabilities, not replace them.

Since the release of ChatGPT in late 2022, I have been increasingly helping clients utilise large language models (LLMs) and other generative AI in the innovation stage of their design journey. With client permission, and in accordance with their policies, AI is a tool that can add great value and enhance our capabilities as designers.

This potential is now starting to be reflected within academic research. Last year, I came across a paper on using AI to ideate from the Mack Institute for Innovation Management at Wharton Business School. The paper titled Ideas Are Dimes A Dozen: Large Language Models For Idea Generation In Innovation tests the ideation capabilities of ChatGPT-4, looking at the productivity of the tool and the quality of the ideas it generates as measured by their commercial potential.

The researchers found that LLMs could increase the productivity of innovation efforts, enhance the quality of ideas produced during ideation, and present novel ideas that are not limited by the self-censorship process most people go through when ideating.

Over the past few months, I’ve had many conversations with clients, researchers and practitioners about generative AI and large language model tools and how they are experimenting with them. Having used these tools in ideation and new venture design processes, it’s clear that while there are great benefits, they are far from eliminating the need for human-powered creativity and ingenuity.

At ThinkPlace, we’re continually uncovering these benefits and learning how to get the best results with clients. By integrating generative AI tools into human-centred design thinking, we can build a robust ideation process.

So, how do we use AI as part of the innovation process at ThinkPlace?

I’m a firm believer in working with technology rather than against it. When we use AI in a way that augments human capabilities, we’re positioning AI as a great counterpart to our designers. It’s not replacing our creative skillset but helps us with additional capabilities such as early research, increasing the number of ideas generated and reducing self-censorship.

At ThinkPlace, AI is helping us in the ways the Mack Institute’s paper suggests. Here are a few ways in which we leverage AI during the design and innovation processes:

To kickstart the research process 
Rather than starting qualitative research from a blank slate, we have started asking ChatGPT to generate a rough framework to build upon. By asking AI to help guide qualitative research through lines of inquiry, research briefs, and ‘how might we’ questions, we can apply design thinking to what it produces and adapt the material to our context and needs.

In the research domain, we can also use AI to create high-level customer personas. When I’m helping clients create personas for interviews, we may start by asking AI to generate personas, which we then refine, sense-check and contextualise within their places and contexts, once we have conducted further research.

As mentioned before, we have done this with clients, taking the output of generative AI tools as a starting point and undertaking further research to build the nuance and subtleties needed. At ThinkPlace, AI has helped us unlock our creativity and critical thinking, as we begin to evaluate, critique and adapt the outputs of the models. This is a useful exercise not just when leveraging AI, in your process, but also when working independently from it. The research paper from the Mack Institute also found that humans were necessary for the idea evaluation and selection process after the AI tool had undertaken ideation. Human skill and judgement are needed to evaluate and understand which ideas have the most chance of commercial success and viability because we can contextualise the nuance of our work a lot better than the models can, unless it is algorithmically bespoke.

The need for human input is further highlighted by the ‘hallucinations’ phenomenon. Generative AI can sometimes present incorrect research and draw untrue conclusions, even going as far as to provide fabricated internal citations and quotes and doing so convincingly (read more on that here). A critical eye, fact-checking skills, and a commitment to using AI to enhance workflow, rather than replace it, are crucial.

Case study: We were working with a group of academics seeking to undertake customer discovery in their research domains to commercialise their research. Through showing some examples of the capabilities of Chat GPT, and providing some example prompts we ran an activity where they generated early personas and lines of enquiries for customer interviews. We then refined these over the coming weeks in coaching sessions as they gained insights from their customer interviews. 

To generate ideas, fast 
The Mack Institute notes that large language models can produce far more ideas than humans can, in a much shorter timeframe. Large language models such as ChatGPT and Bard are great tools for design sprints; by producing large amounts of ideas quickly, they can reduce the workload for innovation teams and unlock creative ideation through refinement and critique exercises

The biggest reason our clients and partner organisations are keen to experiment with AI is a simple case of resource constraints and availability. As global economic conditions have worsened (rising inflation, higher cost of capital, shortage of skilled labour), many companies have reduced spending on innovation and growth projects. By using AI in the ideation phase, organisations can maintain a fast pace in projects, whilst simultaneously turning more resources towards the activities that benefit the most from subject matter experience, such as idea selection and commercial development of ideas.

Case study: We were working with a large government client, running a rapid ideation exercise as part their strategy offsite. After breaking up the room into interdisciplinary groups, and conducting some simple prompt training with ChatGPT, we ran multiple rounds of ideation on ways to decrease biodiversity loss, using generative AI as a counterpart. 

To create room for novel innovation 
One of the hardest parts of the ideation process can be getting groups to put aside their prior knowledge, preconceived notions or heuristics. One of the benefits of AI is that it comes without these, providing us with a foundation of ideas that allows us to think outside our realm of limiting beliefs. While AI isn’t without bias, using it reduces the self-censorship that most of us go through (consciously or not) when ideating. Sure, this will mean that there will be some unreasonable ideas that will never work. But one of those crazy ideas just might.

Because AI doesn’t have contextualised human empathy, real-life experience, creativity or originality, it’s important to have experts include their knowledge when it comes to novel ideas. Often, a novel idea spurred by AI might not work within our complex human systems for a multitude of reasons. Only those with lived experience and design expertise are best placed to evaluate the viability, feasibility and usability of the idea.

Case study: With the same government client, we ran some exercises to combine ideas, and refine them. Two of the ideas selected for further exploration came from initially non-obvious ideas, with the exec remarking that it was likely that they wouldn’t have thought about the ideas if the team hadn’t used generative-AI to generate additional suggestions. These ideas were taken forward for further validation, testing and customer discovery with relevant stakeholders.   

Activity How Gen AI can be used Be aware of…
Research brief Early research synthesis when framing briefs

Starting point to build hypothesis for testing in interviews
Be aware of hallucinations and  bias

Need subject matter expertise to build out nuance
Customer discovery Ideas for less obvious questions which can be used Not all questions will be logical, make sense or be useful
Personas Build a draft persona to nuance and further refine through interviews Bias, may not be useful for very specific or smaller cohorts
‘How Might We’ statements Reduce human preference towards a particular solution through framing of HMW question

Produce wider set of potential HMW questions
Not all generated HMW questions will be logical or useful
Ideation (long list of ideas) Generate greater number of ideas

Generate a wider set of possibilities

Reduce human preference for a particular solution
Not all ideas will be logical or useful

Subject to bias
Idea analysis Build first cut of idea assessment by specified criteria e.g. feasibility, viability desirability framework
Reduce human preference for a particular solution
Need to add layer of human experience and customer discovery results to refine assessment

Subject to hallucinations
Idea combination and links Explore links between ideas and potential combinations of ideas Subject to hallucinations

Ask AI output to include references for cross-checking
Prototyping Generate artefacts for early prototypes e.g. images of products, text for landing pages etc.


So, you want to try out generative AI in innovation? These tools can assist you:

The simplest place to start with using AI as part of the innovation process is to use tools like ChatGPT and Bard as they are. If you want to experiment with visual prompts and input into brainstorming or prototyping, DALL E is also worth experimenting with. Finally, for more incremental product innovation there are tools such as Mixpanel and Zeda which use AI to aggregate customer voice and feedback to inform innovation efforts.

For those starting with AI to augment their innovation process, two of my favourite, easy to use tools for beginners are below. I would suggest starting with these, and then moving onto some of the others above.

Design Sprint Academy – Design Sprint Coach Custom GPT
One category of tools that can be helpful is Custom GPTs, which is a product released by Open AI that allows users to build a custom instance of ChatGPT. One good example of this is the custom GPT that acts as a virtual design sprint coach, built by the Design Sprint Academy in Europe. This is a great way for those who have less experience in the design sprint process to get advice and tools to run their own design sprints.

Board of Innovation – AI Toolbox for Innovators
I’ve also seen other examples, such as The AI Toolbox for Innovators, that utilise an API link to ChatGPT to tailor prompts to ensure they are most useful. These might be used to generate personas, ‘how might we?’ questions and problem statements. Undoubtedly, these are useful as a starting point. However, because of the interface, they can also be limiting compared to the freedom of an open text box offered by Chat GPT or Bard.

In my work, I have seen firsthand the increased speed and efficiency that can come from tools like the design sprint coach and AI toolbox. For those with less confidence using generative AI tools, these are a great starting point.


Where does AI leave us as designers?

While AI doesn’t eliminate biases – there’s wide speculation that its data input is politically biased for one – it can nudge us over the edges of our innovation limits. Humans, whether we like it or not, cannot compute or withstand as much information as AI models can. This is not something to envy or deny, but a truth to utilise if we’re to work harmoniously with generative AI’s expansive benefits. We’re in the caveman era of AI, sure, but that’s further reason to work with its evolution, not against it.

The ethical considerations are a far greater topic that experts in the field are better placed to discuss, but as designers, I think there’s great potential in AI. When we join forces and offer our strong ethical compasses, expertise in design thinking, and real-world experience of engaging diverse voices in problem-solving, we can enhance ideation and bring ideas to light that wouldn’t have emerged otherwise.

If you want to chat about how you can use AI in your innovation process, reach out to



Girotra, K., Meincke, L., Terwiesch, C. and Ulrich, K.T. (2023) ‘Ideas are Dimes a Dozen: Large Language Models for Idea Generation in Innovation’. SSRN, [online] Available at:

NESTA. ‘AI is reinventing the way we invent’. Available at:

MITManagement. ‘When AI Gets It Wrong: Addressing AI Hallucinations and Bias’. Available at:

Baum, J. and Villasenor, J. (2023) ‘The politics of AI: ChatGPT and political bias’, Brookings. Available at: