I gave a talk this summer to a room full of high school leaders. Somewhere in the middle of it I said a sentence I hadn't planned, and it has been rattling around in my head since.
AI does not care where you're going.
I meant it as a warning, but the students heard it as permission, which I think is closer to right. That is not a threat so much as a description of how the thing actually works.
An engine, not a compass
Most of the conversation about AI is about capability. What can it do now, what will it do next year, what will it do to my job. Those are real questions and I spend most of my week on them.
But capability is the wrong first question, because AI is not a direction-setting technology. It is a speed-changing one.
If you're a careful person who checks their work, AI makes you a faster careful person. If you're a person who cuts corners, it makes you a much more productive corner-cutter. If your business treats customers like relationships, automation gives you more hours to spend on them. If your business treats customers like transactions, it gives you a way to process more transactions with less human contact.
Nothing about the tool decides which of those you become. It just amplifies whatever was already true.
Why this is uncomfortable
The uncomfortable part is that amplification is revealing.
A lot of us have been protected by friction. You didn't send the sloppy proposal because writing it would have taken two hours you didn't have. You didn't cut that corner because the shortcut was not available. Your restraint looked like character, and some of it was, but some of it was just the cost of doing the wrong thing being high enough that you did not bother.
Take the friction away and you find out which was which.
I have watched this in my own work. There are weeks where AI has let me serve people better than I could have on my own, and there are weeks where it has mostly helped me stay busy. Same tool, same person, different direction. The difference was never the software.
The pattern of the moment
There is a default setting in technology culture that goes: if we can, we should. It is rarely argued for. It is just the water.
Paul Kingsnorth writes about this better than I can, and I would not call his book comfortable. His argument, roughly, is that we have gotten very good at asking whether something is possible and very bad at asking whether it's good. Progress becomes its own justification. Nobody has to defend it because nobody remembers to question it.
You do not have to agree with everything he concludes to notice that he is describing something real. In AI specifically, the gap between "someone built this" and "everyone uses this" has gotten short enough that there isn't much room left in it for anyone to ask what the thing does to us.
Which means the asking has to happen earlier. Before the tool arrives. Before you are tired and behind and it is right there.
Decide the direction first
The practical version of all this is unglamorous: write down how you will and won't use AI, and do it now, while nothing is at stake.
Ethics written under pressure are not ethics. They are rationalizations with better vocabulary. Every genuinely hard call I have made in this business was easy in the abstract and hard at 4pm on a Thursday with a client waiting. The only thing that helped was having decided beforehand.
At Connexion AI we keep ours short enough to actually remember:
People matter most. When a decision trades human good for efficiency, the human wins.
Values guide decisions, not the other way around. You don't reason from the tool to the ethic.
Responsibility requires accountability. "The AI wrote it" isn't a defense. It never was.
AI should promote human flourishing. Not output. Flourishing. Those come apart more often than you would think.
Yours will look different, and they should. The point isn't my four sentences. The point is that you have four sentences at all, and that you wrote them before you needed them.
What this looks like once it reaches actual systems is more concrete than it sounds. Building automation a person can still be accountable for is mostly a design question, and it has a good answer.
What it's for
Here is the part I actually believe, and it's why I do this work instead of something quieter.
The reason I describe this business as giving people their hours back, rather than making them more productive, is that those are not the same goal and they often pull in opposite directions. Productivity asks what else you could fit in the time. Hours back asks what you would do with the time if nothing had to fill it.
Anthony Hunt puts it better than I've managed to: the point is getting hours of your week back, not to do more, but to do more of what matters most. That only happens if somebody points the tool at it deliberately. Left alone, saved time gets absorbed. It always does.
That is available to almost anyone right now. The gap between having an idea and having the thing has never been smaller. What hasn't changed is that somebody still has to decide the idea is worth building, and for whom, and at what cost to the people nearby.
AI won't do that part. It was never equipped to and never asked to. It is a remarkable tool that has been handed to us, and it's going to accelerate whatever we were already doing.
So the question isn't really what AI can do. It is the older question, the one that was there before any of this:
Where were you going anyway?