Whose Headache Are We Solving?
Last week, my organization held the first in a series of AI share-and-learn sessions. About ten people joined, and the use cases were the kind you’d expect of the early adoption period: drafting communications, turning technical language into something friendlier, working with spreadsheets. All the ways AI can make an individual’s work faster and easier.
As the administrator of our AI account, I can see how often people use AI tools, and what’s interesting is that some people have found these use cases and use AI multiple times each day to help themselves. At the same time, other people in the exact same role aren’t using the tools at all. I can't know for sure, but I suspect the difference isn’t that some people have found advanced uses. Rather, it’s that some people have worked AI into their “normal” workflows whereas others have not.
As I reflected on the share-and-learn session and the differences in adoption, I thought of one colleague’s framework for deciding how to use AI: What headaches do I have? He started with those activities that were frustrating and time-consuming and asked the AI tool how it might help.
What he built from that was pretty cool. Every year, we have to get students set up with usernames and computers. Doing that requires taking information from multiple sources—e.g., what classroom a student is in, where the devices are—and close quality assurance to make sure we don’t accidentally duplicate usernames or assign the same device to multiple people. To ensure that it was done correctly, my colleague approached it as a Friday evening batch process—when he could count on four uninterrupted hours and stable data.
Using the headache framework—in this case, avoiding the late-evening work—he built a tool that pulled the whole process together and did the QA automatically. That alone shaved several hours off the task. But because the tool no longer needed four uninterrupted hours to run, the process doesn’t have to stay a once-a-week batch job. It could happen daily, and with more people able to run it safely, it doesn't have to wait for one person to have uninterrupted time.
The tool was genuinely impressive. But I realized afterward that the headaches framework that prompted the work would only get us so far.
It reminded me of an issue we used to talk about when I worked in banking. If you’ve ever wondered why checks take a day to clear, or why there’s a cutoff time each day for sending a wire, it’s because the whole system is built upon overnight batch processing. Historically, bankers met in person at the end of the day to settle checks and what each bank owed the others. And even once it was electronic, it was easier to run the system overnight when there was more computing capacity.
It’s a terrible experience for customers who need their money immediately or who need to know right now whether they have enough money to spend at the grocery store. But because the overnight batch processing worked fine for the banks and the bankers—people with enough money not to notice a day's delay—it wasn’t enough of a headache for them to fix the headache it caused for customers.
That’s how I realized that headaches—a personal problem to solve—might be a great way to help people get past the inertia of integrating AI into their work process, but it might not help us get to the solutions we ultimately want—the ones for the people we're doing the work for, not just ourselves.
An AI tool that automatically sends a report saves me from having to compile and email it myself, but it’s not the same as building a data product—i.e., something formatted and organized so the person on the receiving end can understand it and act on the data. One version solves my headache; the other solves the recipient’s headache.
I could prompt an AI tool, “create a neatly designed poster listing classrooms and students so parents can see where to go,” to save me the hassle of formatting it myself. However, if I were solving for the parent being able to find their way to the classroom, I’d identify where the poster would hang and add information like, “Go to your right, turn left at the corner, and look for this logo on the door.” But that second step is extra work, so it may not get done unless I’ve included the parents’ headaches in my definition of success.
For our student tech setup, it'd be easy to stop at the current solution, which delivers significant time savings. But if everyone had a vision for the outcome that parents and students might want—real-time, seamless enrollment—we would all go the extra mile, scrutinizing all the human coordination processes beyond the tool to actually deliver it. Just solving for one person’s or one team’s process wouldn’t get us there.
It’s like a relay race. Solving internal headaches may be the first leg of AI adoption since there’s something in it for each employee. But at some point, the baton must be passed to a vision that defines the finish line—the outcomes that make the effort worth it for the organization. Those are bigger than individuals’ work, and they’re probably found in the headaches of the people we’re doing the work for, not our own.