Begin with context.
Here I am. Here’s what I do. Here’s what I despise. Here’s what I want to keep doing. Here is my workflow. Here is my chaos.
You don’t need to arrive with the right question.
Tell me what happens. Show me what you’re working with. Explain the part that makes you sigh before you’ve even opened the application.
And tell me what you love. That deserves a place in the conversation, too.
If we begin with “What can we automate?” we’ve already narrowed our attention to the work as it currently exists. We may become excellent at automating something nobody needs to do anymore.
An impressive achievement. For some reason, we still have the problem.
Context gives us more to work with.
A recurring task might belong to a process that has outlived its purpose. Something you hate doing might be keeping you from something you’re particularly good at. A project that seems unrelated might already contain part of the answer.
Bring the life and work the assistance is supposed to serve.
That can be wonderfully ordinary material.
A list of supplies and a need. A few examples of color from Egon Schiele. Three attempts at explaining an idea. A spreadsheet with an unfortunate family history.
“Here’s what I have. Here’s what I’ve tried. Something about this interests me.”
We can begin there.
The frustration belongs in the conversation. If you dread a task, keep finding ways around it, or spend your energy compensating for something that never quite works, that tells us where to look.
What is making this so difficult? Where does the trouble begin? What have you learned to tolerate because it has always been there?
Pain is information. Give it some attention.
Give the pleasure some attention, too. What holds your interest? What would you happily spend more time doing if the other stuff would kindly get out of the way?
Those answers help us understand what a useful change would make room for.
In my own work with an AI assistant, keeping the details in view gives me room to imagine. The materials, the purpose, the constraints. We can follow a possibility, then check it against what we actually have.
Sometimes we need to bring a detail back into view or correct an assumption. The context develops as the work does.
And connections emerge.
Work on one project. Bring another into the conversation. Notice a similar need appearing somewhere else.
Seven projects may begin to reveal one larger idea.
That’s how Surface began taking shape. Making art, researching funding, developing products, managing information: activities that looked separate kept revealing connections. The same details were needed in several places. One decision changed what should happen elsewhere.
I couldn’t have handed over a complete specification at the beginning. I hadn’t discovered it yet.
You’re allowed to discover what you’re making while you make it.
Give that discovery something concrete to work on. A sample. One working page. A version someone can try.
Then pay attention to what happens—including how the work feels.
There is effort involved in making things. There is also the afternoon when you’re beating your head against an idea and getting increasingly elaborate versions of the same wrong result.
Stop.
With or without an AI assistant, you’re allowed to reassess. Have we misunderstood the need? Chosen an awkward approach? Become attached to something because we’ve already spent six hours on it?
AI can produce another attempt very quickly. It can help you spend the rest of the evening in precisely the wrong direction.
Ask what the failed attempts are telling you before requesting another.
You might need a different method, a smaller experiment, a conversation with someone who knows something you don’t. You might decide the idea can wait. You might leave it behind.
Let an attempt fail without requiring the person to become a failure.
Keep what you learned. Give your next decision the benefit of it.
At work, people need room to do this openly. Time to explore, help when they get stuck, and permission to say something isn’t working. Otherwise, you’ll get a demonstration polished enough to conceal everything you needed to learn.
The next worthwhile thing may be difficult to see from the existing task list. Understanding the person, following the work, and noticing what happens gives it a chance to emerge.
And if you’re bringing your own useful mess, please don’t tidy all the life out of it before we meet.
The interests, frustrations, unfinished attempts, and apparently unrelated projects give us something to work with.
Here you are. Let’s start there.