It's time we stopped asking for vases.

My first interactions with AI came in 2015 when I joined IBM. Watson was all the rage, but it was still early days for the technology. Despite the tens of millions of dollars poured into marketing these incredible capabilities, Watson wasn't ready for prime time. I came to learn that the singular entity was actually an array of APIs that were focused on one ability, like compare and contrast. For the sake of conversation, each ability had a grammar school grade level. As I recall, the compare and contrast part of Watson was "in the sixth grade" when I was there. The people I knew who created things with Watson had to put in a lot of effort just to get the technology to do basic things like suggest a type of beer or cocktail based on answers to a handful of questions. Because of this, I didn't see the benefits of this technology aside from using it for the sake of using it.

When generative AI was made available to the public I, like you, bore witness to an insane barrage of crap posted to every nook and cranny of the Internet. It was interesting to see how far AI capabilities had advanced, but that curiosity lasted for about a day. Prompt engineering became the latest must-have skill. Many folks I know who advocated "hand crafted" quality in digital work were duct taping AI to their services and thought leadership. As AI trends shifted so did the language on their company services pages and the topics of their blog posts. And yet I didn't see anyone creating anything radically better from what they did before. Once again, I didn't see a good use case for using AI for anything other than cheap tricks and trying to remain relevant to the winds of zeitgeist.

It wasn't until about a year ago with the release of a new model for ChatGPT that I started to see something different. I'm not an engineer so I can't tell you what changed technically but the responses I got to questions were smarter. Instead of interacting with Captain Obvious, the new version offered actual, interesting insights. As it continued to improve and respond to my ongoing curiosity I began to realize that AI was turning into more than a useful tool. It was becoming a useful collaborator.

Orientation at IBM Design took place in a specially designed studio with a track system installed in the ceiling. Eight-foot-tall whiteboards were mounted to the track in a way that made it possible to move them around to form different sizes of work spaces. All the tables and chairs had wheels to accommodate instant reconfiguration of the space depending on the need. It was a cathedral for design unlike any that I had seen before or after working there.

It was in the "blue space" that Doug Powell kicked off orientation for my new employee cohort. We started the day with a simple exercise.

"Spend thirty seconds and draw a flower vase," he instructed. We reached for a pad of 3M Post-It Super Sticky Notes from the Playful Primaries Collection and a Fine-Point Sharpie and drew an image as instructed. At the end of the time Doug didn't wait for some folks to stop. He advanced his Keynote presentation to a photo of a person holding up a drawing of a flower vase that was similar to what everyone in the room had just drawn.

Doug went on to say that we all drew the same thing because the instructions or the ask prescribed the solution. This is why, he argued, so many companies are not able to genuinely innovate. Instead of taking a step back to reconsider everything, they end up creating just another version of the thing that came before it.

The presentation advanced again along with a new instruction. "Now I want you to spend two minutes creating an experience for enjoying flowers." When the timer whistled for us to stop, half of the cohort was rushing to finish drawing their ideas. As we went around the room to present our ideas it was clear that removing the solution from the ask radically changes the possibilities and ideas. This simple act affords trust and gives permission to explore which leads to way more interesting ideas and outcomes, and ultimately innovation.

You get what you ask for, and so many times the ask includes heavy constraint. This is why I don't understand why so many people are focused on creating the perfect, yet generic, universal prompt for AI. Sure, there are times when including instructions, conditions, and constraints are useful but I've found that too much turns AI back into Captain Obvious. Yet, from my observations and conversations this is how most people are using it.

In short, we're using this technology to improve what we've come up with on our own instead of using it to explore new ideas and perspectives that we can graft onto ours which will always lead to learning and creating outcomes of higher quality. I think that means changing how we think about AI not as a tool but a collaborative partner who can bring the full weight and might of human knowledge, thinking frameworks, research methods, and enhanced synthesis to every user.

Whereas I used to see AI as a gimmick and then a gimmick factory, I now see it as a game changing technology that, when used correctly, can and should advance our knowledge, thinking, skills, and ability to deliver beyond what we were capable of prior to these advancements. AI is not your friend, but it is a tremendous competitive advantage bringing warp speed to curiosity and the curious.

For the last couple of months I've been working on something that I'm genuinely excited about—a book that explores this different perspective on AI collaboration, paired with a playbook that turns these ideas into something you can actually use. I'm not interested in adding to the noise about prompt engineering or contributing to the AI hype cycle. This is something different called Stop Using AI Like A Vending Machine that helps people move beyond using the technology transactionally and start using it as a thinking partner that makes your work genuinely better, not just faster.

When everyone has access to the same AI tools, the real advantage isn't going to the people with the fanciest prompts—it's going to the people who know how to think with machines, not just through them. I think this might actually help people, and I can't wait to share it.