AI and Design Thinking: A Marketing Problem Example
There is a lot of hype about AI, but can it help with design thinking? If so, how? Here are our takeaways from our experimentation with 10 AI mind mapping and idea generation tools. We won’t mention which ones we used, as some are still developing, but our findings are sufficient to provide an idea of what to expect for now.
To test how well AI worked for creating a business solution, we partnered with www.Gozeeit.com, an event marketing company. The problem we wanted to solve was how to improve current marketing efforts. We needed to get the word out to users, both people who post events and viewers looking for things to do.
In our test, we wanted to see how the AI systems compared against traditional human results we’ve experienced in design thinking seminars. We started the experiment by asking, “Give me some marketing ideas for an online event promotion company.” We modified the question a few times and then narrowed the prompts down in response to the answers provided. For instance, if the response was “Use SEO,” we asked it how. Without going into graphic detail on our process, here are our abridged findings.
Findings
Finding 1
Nine of the 10 systems had the same or similar ideas from the first and second prompts. They were presented in different orders, with some being mapped and others creating lists. Many of these ideas were already being used by www.gozeeit.com, while others were either not appropriate or required additional background work before they could be implemented as they were too general. This is common in the first rounds of design thinking ideation with a new team. What was surprising is that 90% of the systems essentially said the same thing—ideas that were already common.
Finding 2
None of the tools provided anything novel or innovative. For instance, if we were looking for Red Bull or Virgin Company-like marketing stunt ideas, AI tools provided a list of what these companies did, but the ideas were neither specific nor new. What the prompt did reveal was a list of attention-grabbing marketing ideas that have already been done and could be built on or modified by humans in the loop.
Finding 3
The responses were standard responses without details or examples. Writing a series of increasingly specific prompts was required to garner more details. The results came back as being methods that were already used elsewhere. There appeared to be little machine learning or ideas being connected.
Finding 4
Even with repeated prompts to narrow down some subjects (e.g., “Tell me how to improve paid social media advertising”), the results provided general ideas that were very common and not specific enough to be useful, innovative, or unique.
What AI Lacked
Although some AI tools helped create maps, we generally found that the AI tools, including those design specifically for design thinking or mind mapping were the same as having a generic (as opposed to a specialist) person listing off ideas. The AI systems acted more like search engines, showing few to no references, examples, or case studies and little in the way of innovative idea creation or connecting ideas from different realms together. To get concrete examples, we had to already have an understanding of examples from specific companies. For example, we had to mention Red Bull by name to get more concrete examples. AI would not have produced those examples if we had not known about Red Bull’s marketing efforts already. We’ll mention more on this later.
Outrageous marketing ideas were absent, and there were no “how to” lessons unless we referenced specific outrageous examples on which the system could focus. All the information found was already taught in college marketing classes and run-of-them-mill advertising agencies—just with less detail in many examples.
Our Recommendations and Observations
Recommendation 1
AI can fill in gaps during a brainstorming session that may simply be forgotten. Having someone in the room to run AI queries can help look at ideas or categorize them while filling these gaps. The benefit is that common ideas are not forgotten so they can be used to build upon for unique ideas. The downside is that others may become wedded to familiar ideas. Don’t expect innovative, groundbreaking, or blue ocean ideas.
Recommendation 2
Only one AI system is needed. The results from across the 10 systems were fairly equal. A few made interesting diagrams, and others made lists and notes. Some purported to be the best system because the ideas could be moved around like sticky notes or tied to other ideas. While this neat feature can work for online collaboration, it is not AI and was sometimes more awkward than moving sticky notes in person.
Recommendation 3
Following on from another study we were engaged with previously, AI is only as good as the prompts it receives. We tried narrowing down the scope of the responses, and that certainly helped. However, from our prior experience and the level of detail desired, a good prompt could be either a series of narrowing questions or one prompt of 300 words. Since we were not doing academic research, the chances of finding false information were low. When we compared what we found against suggestions by a marketing firm, many were the same. When we compared results against specialists, the specialists were able to provide alternate ideas and explain why those ideas would be good in different situations (e.g., backend website linking, which AI did not produce as but one example).
Again, nothing was innovative; it was just standard stuff. That said, standard stuff is suitable for people with little marketing experience or as a starting point to generate ideas, provided design participants don’t become fixated on those ideas alone.
Recommendation 4
We found that pen to paper as a group was much more fulfilling than simply looking for answers from a computer. The team engagement, riffing off each other, uniqueness of ideas, connections between ideas, specialist explanations, making connections between disparate backgrounds, and communication is far better in almost all our design thinking seminars compared to what AI could do. The role of specialists, in particular, provided background explanations on thinking that could be riffed off by other participants, resulting in unique ties that AI did not offer.
Observation 1
A characteristic of AI is learning, as opposed to software, which spits out results based on programming. With the particular problems we were throwing at it, it did not appear to learn as much as just list ideas. This likely resulted in all systems arriving at similar results lacking imagination.
Observation 2
AI was simply not creative for this problem set. Unlike AI music or art, which can create nifty images (if still problematic with proportions and realism), our design thinking creative needs were largely unmet. AI could not tie concepts from two different realms to create something unique. In this sense, AI was doing the job of software.
Additional Information
We recently attended an e-commerce conference in Europe. A keynote speaker from a major consulting firm noted that the rise of AI will see all the data on the internet skimmed within 24 months to help with the learning models. This results in a couple of issues. First, new ideas are likely to level off unless AI can learn faster than it is right now, including drawing from many different specialist backgrounds, such as a design thinking team can do. Second, unless AI can learn to create unique solutions from a foundation of knowledge, the use of humans in the loop will likely produce better results. Third, AI does not appear to be able to link different backgrounds together. For example, connecting baking molasses with snow removal is something AI cannot seem to fathom.
Summary
AI can be a tool to help fill gaps in listing general ideas or for someone with little marketing experience to see a series of topics that affect marketing. But at this stage of AI, new, crazy, and innovative ideas do not appear to be a strength of any of the systems we tested. AI worked to create lists of starting points for marketing ideas that have already been used. This was helpful if the prompts can be written with background knowledge of innovative marketing examples. To that end, human creativity still appears to best the machines for creativity and innovative marketing ideas while AI plays a supportive, base role.