The Two AI Skills I Would Put First: Imagination and Creativity

In one of my companies, our customer success team would randomly call clients to find out whether there were problems. We had thousands of client interactions a month, and the volume stood in the way of having someone review meetings for signs of trouble.

That is the kind of business problem I have in mind when I say the two AI skills I would put first right now are imagination and creativity.

By imagination, I mean seeing something your business could do that you hadn't considered possible. Creativity is connecting the information, the tools and the people who can help you try it.

We started by asking how to prevent churn

We started with a question: how do we prevent churn? We wanted to help our customer success team identify problems early, and looking for signs in client meetings was the obvious choice.

The team was already reaching out to clients, but the volume made it difficult to review meetings for signs of trouble too. So we looked at whether AI could help identify meetings that needed attention.

A client complaining, asking for the same information repeatedly, pointing out a missing report or something that hadn't been delivered, those were things we wanted to notice. So was a bad or upset tone in the conversation.

We used AI to classify the meetings for signs like those. If it flagged a meeting as concerning, we would send the transcript to the customer support team, and they would handle it from there.

The transcript gave them the conversation to work with. The AI's classification wasn't the response to the client, it was a way to bring a meeting to the attention of people who could do something about it.

I am describing how we approached the problem, not claiming a measured reduction in churn from that system.

AI classification gave us a way to test whether we could help the team find problems in that volume of conversations. The support team still handled what needed to happen with the client.

I've tested other things too, lead filtering and classification, monitoring messages to make suggestions. These are the kinds of things I mean when I talk about creativity in a business, looking at work we already do and finding another way to help with a problem in it.

Imagine what would solve the problem before choosing a tool

If you don't think of yourself as creative, I would start with the problems you already understand. You don't need an unusual idea to begin looking at why work gets stuck or why clients leave.

Look at where you spend money with little return, the work people dislike doing, what clients keep asking for and what they don't want anymore. There is plenty to think about inside an existing business before you go looking for an AI use case somewhere else.

Then spend time understanding why the problem happens. A client repeatedly asking for a report gives you somewhere to look, but you still need to find out whether the report wasn't delivered, wasn't clear or didn't answer what they needed.

Those are different problems, and I would want to understand which one I was trying to solve before asking someone to build anything.

From there, imagine what the best solution would look like in a perfect world. Leave the technology out of it for a moment and think about what you would want to happen for the client and the people doing the work.

And please don't turn this into telling your team they need to be more creative with AI while leaving them to guess what matters. They can help identify problems and suggest solutions, but you still need to decide which problems deserve the company's attention and give people room to work on them.

Learn enough about AI to see what you could try

The meeting example is why I care about learning what AI can do beyond writing an answer in a chat window. Once you understand that it can look for signs of trouble in a conversation, you can consider how that might help a team with more conversations than someone can review.

That doesn't tell you whether it will recognize the right problems or produce something the team can use. It gives you a possibility worth testing, and the more I understand about what agents can do, the more of those possibilities I can imagine.

We are still figuring out how to apply this technology in our businesses. I would keep learning and trying things rather than assume a fixed list of tasks covers what will be useful next.

That is why I think owners need to experiment with AI personally. In Don't Be a Detached CEO, I explain why that understanding matters even when someone else handles the technical work.

You can research what other people are doing, try things you want to understand, and pay attention to what the tool handles and what still needs your help. You are building enough familiarity to connect a business problem with something worth investigating.

Not every personal experiment needs a business case. Learning what a tool does and deciding to put it into the company's work are different decisions, and you need room for the first before you can make the second well.

Now, there are useful skills we can teach. The thread that prompted this article includes prompting, fact-checking, working with your documents and knowing when not to use AI. I agree those are worth learning.

We can teach methods and safeguards too. What no fixed checklist can do is tell you which useful possibilities to pursue in your particular business, because that requires understanding your problems and testing ideas against the work.

And when AI does free time in people's calendars, I would make room for some of that exploration. Decide which problems you want people to investigate rather than assuming a faster way to do their existing work will lead them there on its own.

You can get help turning the idea into something that works

You don't have to do all of the creative work yourself either. You can bring the problem and the possibilities you've imagined to more technical people and ask them to help you create a solution.

Give them enough to understand what you are trying to change, why you think the problem happens and what a better result would look like. That gives them something to work through with you, rather than a request to put AI somewhere in the business.

They can help you understand what is possible, suggest an approach you hadn't considered and work out how to connect the tools and information. You still need the people who do the work involved, because they are the ones who can tell you whether what comes back helps.

For an experiment like ours, I would ask the team to look at the flagged transcripts and tell us whether they contain problems they can act on. I would also want them to point out concerning meetings the system missed.

That feedback gives you something specific to change. You can investigate why a complaint was missed or why a meeting was flagged when the team didn't see a problem, then try again.

And flagging the right meetings still doesn't tell you whether fewer clients are leaving. Someone has to act on the information, and you need to look at the business outcome you wanted to improve before calling the experiment a success.

For the problem you choose, bring the outcome you've imagined to people who understand the work and the technology. Agree what a trial needs to show, then ask the people using the result whether it helped them do what you intended.

Use what you learn to decide whether to improve the approach, try another one or leave it alone.