What AI makes possible—and what it asks of us.
The work often begins with something waiting. A report needs retrieving. Copy needs to make its way into a design. A flyer needs an image, and the lead designer is already booked.
Those are some of the bottlenecks I’ve been helping one client clear. Each starts with understanding where the work is waiting and what would allow it to move.
For that same client, I built an AI-learning series with nine short episodes. Each came with a prompt pack and a manager’s discussion sheet. There was also a browser-based Playbook, somewhere to return when the video was over and an actual piece of work was waiting.
I designed the series to give employees a foundation for working with the AI tools their company had approved. The films introduced ideas, the prompt packs provided a way to try them, and the manager sheets made room to question the results.
That is where my consulting and creative work meet. I like making things that catch your attention. I care just as much about what happens after I’ve caught it.
A convincing first result can arrive before we have properly decided what we want from it. The draft is on the screen, the image looks finished, the explanation has a lovely rhythm. There is a natural temptation to judge the work at the moment it becomes impressive.
I try to stay with it a little longer.
Who needs this? What are they trying to do? Which parts can they trust, and what would help them notice when they shouldn’t? Those questions shape the brief, the material I make, and the way I judge whether it is ready to leave my hands. They also guide the system work I do in my own practice.
When I return to a long project, I need to know what we decided, why we decided it, and what remains unfinished. A polished summary can send the work in the wrong direction if it quietly turns a suggestion into a decision.
I’m building and testing a personal AI system around problems like this. The continuity work has helped me return to a project with decisions and their supporting records in reach. Testing has also exposed cases where a check succeeded but the action that followed still went wrong.
Those findings have changed what I am willing to rely on. They have also made me reexamine my own explanations. A failure can be real while the explanation I first give for it is incomplete. Keeping the evidence matters because, eventually, I may need it to argue with a sentence I rather liked.
What I learn in that work changes how I approach the next problem. Applying it well still means understanding the person, the task, and the circumstances.
The working standard is much like cooking. A recipe gives you a starting point, but the ingredients and the occasion give it meaning. A meal can be prepared with the care of a Michelin-starred kitchen and still be plain wrong for the person in front of you. Attention to those particulars is part of the craft. So is understanding when and why to alter your approach.
That is the kind of expertise I bring to AI work: practical enough to produce something meaningful, careful enough to examine it against what it is supposed to achieve, and willing to revise a judgment when the evidence points to a better approach. Learning is part of doing that work well.
The larger question is what these systems make possible for the person using them. More room to think? A clearer way to learn? The ability to take on work that previously felt out of reach? That is what makes AI worth my attention, and it gives me a more demanding question to ask than how quickly a model can produce an answer.
If an hour of drafting becomes an hour of prompting, checking, and repair, the time hasn’t been saved. Whether that exchange is worthwhile depends on what you have at the end. Did you reach a better result? Did you understand something you couldn’t make sense of before? Or have you acquired another process to manage, with more material to sort through and the original problem still waiting?
Sometimes the same hour, spent differently, is a real gain. A rough draft can give someone a way into writing they had been putting off. Several visual directions can help a team discover what it is trying to say. Help working through a difficult document can make an informed decision possible.
Time saved matters too. Retrieving a report is a useful task to automate precisely because the person who needs it has other work to do. Automating retrieval should leave the person with less to do once checking and upkeep are included. When checking and repair consume the benefit, the approach needs to change.
This is where I put the question of value: what does the person gain across the whole task? The answer might be time, a better result, a clearer understanding, or the ability to do something they could not previously attempt. Producing more material gives us none of those answers by itself. I want to see what someone can actually do with it.
That is the ambition behind the tools I’m building. They should grow with the people using them, carry useful experience forward, and make the next thing easier to take on. The measure of that work is in someone’s life: a project they can finally finish, an idea they can explore, or an evening they can have back.
Leave a Reply