Brainstorming with ChatGPT: creativity with limits

“Using the brain to storm a problem” – this is how Alex Faickney Osborn described his creativity technique. Acting smart to find a solution to a problem – countless organisations have used brainstorming in workshops and meetings to collect ideas, develop new products or devise strategies. It’s no coincidence that it is known as a classic creativity technique.

In practice, however, it often turns out that brainstorming sessions do not always lead to the best possible results if they are not optimised and the rules of exchange are not adhered to. This is where ChatGPT comes into play. “Using ChatGPT to storm a problem” is already a reality in many organisations. In which situations is the popular chatbot a good brainstorming partner and where does it fall short? Time for a situation analysis and thematic preparation.

Brainstorming process and rules in a nutshell

There are many good descriptions of the brainstorming process and rules on the Internet. Here is a brief summary:

Brainstorming follows a clear sequence: preparation, realisation with idea generation and evaluation, and follow-up. In the preparation phase, the appropriate team for the problem is put together. At the beginning of the process, the rules and time frame are discussed, a minute-taker is appointed and the problem is described by the moderator. This is followed by the collection and documentation of ideas by the participants and then the evaluation of ideas through grouping, analysis and prioritisation. As part of the follow-up, the results are processed and made available to the participants – with appropriate tasks if necessary.

The rules ensure respectful, open communication and ideally promote innovative solutions. There must be no criticism of the ideas or the idea providers. Every idea, no matter how unlikely, is welcome. Ideas can be developed further by everyone. There are clear time limits and the objectives of the exchange are communicated. And: Participants can contribute, but do not have to.

Typical problems with classic brainstorming and obvious advantages of ChatGPT

In practice, similar problems can be observed in classic brainstorming across organisations:

One common problem is the dominance of individual people. Extroverted and vocal participants tend to dominate the discussion, while quieter and introverted people often remain in the background and are unable to contribute their ideas. This leads to a one-sided spectrum of ideas and can lead to valuable contributions being ignored. [1]

Another problem is the early criticism of ideas. Suggestions are often evaluated as soon as they are made, which inhibits creative collaboration and means that some participants do not dare to express unusual or risky ideas. This critical attitude can severely restrict innovation.

A lack of structure during the meeting can also be problematic. Without a clear process and targeted moderation, brainstorming sessions can easily become chaotic, causing the discussion to digress from the topic and valuable time to be lost. A structured approach is therefore important in order to maintain focus and organise the brainstorming process efficiently.

Another common problem is the lack of follow-up on ideas. Many good suggestions are often made during the brainstorming session, but these are not followed up or implemented. This can be frustrating for some participants and can lead to a lack of creativity and commitment in future sessions.

Time pressure can also affect the quality of ideas. When brainstorming sessions take place under time pressure, participants can be rushed and make less thoughtful suggestions. Adequate time planning is therefore important to give participants enough space for creative thinking.

Unclear objectives and inadequate preparation often lead to results that do not meet expectations. If the objectives of the brainstorming session are not clearly defined or the participants are not adequately prepared, this can greatly reduce the effectiveness of the brainstorming session. Clear communication of the objectives and careful preparation are therefore essential.

Finally, a lack of documentation and follow-up of the ideas can result in no concrete measures emerging from the brainstorming session. Without systematic recording and follow-up of the suggestions, there is a risk that good ideas will be lost and the meeting will not produce sustainable results. Thorough documentation and a clear plan for implementing the ideas are therefore crucial to the success of a brainstorming session.

Of course, these problems do not occur in every classic brainstorming session, and of course many of these problems can be solved by human intervention: For example, a moderator can immediately stop criticism of ideas or idea providers and point out the rules that have been discussed. Or participants can address the resulting time pressure in order to agree on a better framework for brainstorming. But how does ChatGPT respond to the problems described?

• ChatGPT knows no time pressure. The chatbot delivers answers to prompts in a matter of seconds. As soon as the enter key is pressed, the answers are built up word by word; even after many months of use, it’s nice to look at.
• ChatGPT is a machine and as such is neither introverted nor extroverted. It provides perspectives without pushing itself to the fore and without worrying about being the centre of attention.
• ChatGPT is neutral and only criticises ideas when prompted. The chatbot is unbiased and works without prejudice or personal sensitivities. It knows no envy and also behaves neutrally when it comes to the further development of ideas.
• ChatGPT can help to formulate clear goals and a structure in preparation for the brainstorming session. If this information is made available to the participants in advance, this increases the chances of a good exchange.
• ChatGPT documents ideas on demand so that no valuable suggestions are lost. Follow-up is also made easier.

Unsurprisingly, ChatGPT offers some obvious advantages. But are there also disadvantages?

Scenarios for using ChatGPT in brainstorming and possible disadvantages

Before I go into possible disadvantages, I would like to briefly highlight the scenarios in which ChatGPT can be used for brainstorming:

• In a one-on-one work situation, ChatGPT can serve as a personal sparring partner by generating ideas and providing suggestions that stimulate the user’s creativity. For example, it is easy to find alternative headlines for blog posts or structure content for articles. [2]
• In a team of 2 or 3 people, ChatGPT can serve as an additional source of input. It can provide impulses and suggestions that can be discussed and further developed within the team. [3]
• In principle, it is also conceivable to use the chatbot selectively in workshops or meetings, but there is a risk that the answers will be consciously or unconsciously rated higher. There is a risk of a reverse source argument, so to speak: The idea comes from ChatGPT, ChatGPT is an AI-supported chatbot, the idea sounds plausible, so it must be a good idea.

In the course of a typical brainstorming session, these scenarios belong in the realisation phase. But what about the preparation and follow-up phases?

In preparation for a brainstorming session, the tool can provide relevant information and current trends on a specific topic. If this information is made available to the participants in advance, they can better prepare for the meeting. In addition, ChatGPT can also help with the planning of the brainstorming session by making suggestions for the agenda and schedule to make the session more efficient. It can also make suggestions for different brainstorming methods and techniques that can be customised to the team’s specific needs and goals. [4]

It is also conceivable that ChatGPT generates a list of initial ideas or questions that serves as a starting point for the discussions. Or that different perspectives are shown that the participants can include in their considerations. Going a little further, recommendations for role allocations would even be conceivable, in which individual employees could, for example, take on the perspective of customers and others that of suppliers.

In addition to collecting ideas, the brainstorming phase also includes evaluating ideas. Here, the chatbot can provide support in grouping the ideas. Thanks to its ability to analyse text and recognise patterns, it can identify similar or related ideas and make suggestions for categorisation. This facilitates the organisation of ideas, promotes clarity and supports the process of evaluating and implementing ideas. The chatbot can help to categorise ideas into thematic clusters, which simplifies further analysis and decision-making.

And how does ChatGPT support follow-up work? The tool can help to create a detailed record of the meeting by summarising the ideas and discussion results. In fact, that’s all there is to it, but if you ask ChatGPT in a one-to-one situation what else it can do for follow-up, the tool fantasises about creating resource plans, designing action plans and assigning responsibilities. But where does the relevant information come from? How does the tool know, for example, which employee is available with which skills and which capacities?

Indirectly, ChatGPT also provides information on the disadvantages of using it:

• Although ChatGPT has been trained with an incredible amount of data, the tool lacks specific knowledge about the internal processes and details of a project or product development. It lacks knowledge of the company’s DNA and a feeling for what is feasible and what is impossible. As simple prompting is often not enough to convey the relevant context, it is advisable to work with your own GPTs if necessary
• Just because an answer from ChatGPT sounds plausible does not mean that it can be implemented or makes sense in practice. The AI has no practical experience. Of course, it also accepts no liability in the event that an idea cannot be realised.
• ChatGPT cannot capture the emotional dynamics and interpersonal nuances of a team that are important for effective brainstorming. The look on a colleague’s face, the approving smile of a colleague, the spontaneous development of another colleague’s idea, the situational drift – ChatGPT does not and cannot recognise any of this. [6]

In discussions, the issue of persistence is often seen as a disadvantage. ChatGPT cannot store information between sessions. However, it is foreseeable that this will soon change, especially as previous sessions can be recalled in GPT-4o, for example. [7]

Conclusion: The tool lacks real knowledge for in-depth brainstorming. And the advantage of being a machine and not a human, i.e. not “suffering” from human sensitivities, can prove to be a disadvantage in other respects: Empathy is an important ingredient for brainstorming and collaboration in organisations. It is a driving force for tackling projects together with vigour and enthusiasm; it provides the energy to try out new things together.

The real benefits of ChatGPT for brainstorming

ChatGPT is fast and provides an amazing amount of ideas on virtually any topic imaginable. At the same time, however, on closer inspection it lacks the detailed knowledge and experience needed to generate really good ideas. So far, so good. But what is the actual benefit of ChatGPT for brainstorming?

From my point of view, the main benefits are fourfold:

1. ChatGPT is a tool for beginners and professionals. For beginners, the tool can provide a wealth of inspiration and basic information at the touch of a button, making it easier to get started with a topic and providing structure and orientation. However, many professionals are not looking for an endless list, but for a single impulse that will help them in their context. They are able to transfer part of a statement to their situation, combine this part with other information and take it a step further. In this way, they can use ChatGPT to overcome a phase of stagnation and use the tool as a creative stimulus for thought.
2. ChatGPT enables ideas to be continuously deepened. Through targeted prompting, an idea can be refined over several iterations, leading to deeper and more thoughtful insights. While some people may eventually see repeated prompting as criticism, the chatbot doesn’t care how many times it is asked about a topic.
3. And ChatGPT can help to look at problems from different angles. People often find it difficult to leave their usual way of thinking. ChatGPT can actively support a change of perspective with targeted questions or even look at a topic from a certain perspective. Both are fantastic opportunities!

And the fourth point: the real benefit, the magic of brainstorming comes from the interaction between AI and the individual. ChatGPT offers creativity with limits. At the touch of a button. Knowing these limits and shifting them selectively is your and my job. This is how we develop AI forward, how we make progress in a specific context and perhaps also how we develop our expertise.

Extra bonus

Here you will find 3 additional questions about brainstorming with ChatGPT answered by Michael Schenkel (please press plus):

[1] Katharina Boguslawski has written an interesting article about introverts in brainstorming sessions, in which she gives various tips on how to utilise the creativity of introverts.

[2] Even though ChatGPT is often used to generate texts, the linguistic quality of the texts is not yet top-notch. This is indirectly proven by tools such as ZeroGPT, which claim to be able to reliably recognise texts from ChatGPT. You are welcome to see for yourself, but beware: even these tools have not yet reached the end of their development.

[3] Dr Andrea Herrmann has published an article on 7 IT use cases with Chat GPT that is well worth reading.

[4] Here you can find a list of variants and alternatives to classic brainstorming.

[5] The problem can be solved by organisations training their own GPTs with their own data. This is currently still relatively time-consuming, but it is foreseeable that the training time for corresponding chatbots will be drastically reduced in the future.

[6] The number of AI-supported tools that can provide emotional, empathetic or humorous responses – increasingly also verbally – to make interaction with the tools appear more human is growing steadily.

[7] Interestingly, it can even be observed that GPT cannot forget some instructions. An instruction that often works between humans, such as “Forget that for a moment, I would like to discuss XY for a moment”, does not really work in dialogue with the chatbot.

Ethical and data protection issues in connection with the use of AI-based chatbots are also repeatedly raised in discussions. Certainly an exciting topic, but one for another blog post.

If you like the article or would like to discuss it, please feel free to share it in your network.

Michael Schenkel has published more articles in the t2informatik Blog, including:

t2informatik develops and modernises software as a service provider. Click here to learn more.