There are moments when the future arrives not with a trumpet blast, but with an absolutely terrible word.

Agenticness.

There it is. I apologize.

I wish I could promise you will never have to hear it again, but I suspect the opposite is true. You may hear agenticness much more often than you would like to admit. It has all the elegance of something invented by a committee that had already been meeting too long. Nevertheless, apparently we need it.

And, annoyingly, it describes something important.

I have been trying to understand the difference between an artificial intelligence that answers a question and one that actually does something. Until recently, most of my experience with AI began with a prompt and ended with an answer. I asked. Molly responded. Even when the response was extraordinary, I remained responsible for whatever happened next.

That is changing.

Imagine that Molly and I have been talking and discover something worth writing. Once I approve it, I could tell her to make an audio version, open the website, create a new post, add the text, upload the audio, check that everything looks right, and publish it.

That is a fundamentally different relationship with a computer.

I haven’t explained every mouse click. I have described an outcome. The artificial intelligence has to figure out the intermediate steps, use different tools, encounter whatever it finds along the way, make small decisions, and continue toward the goal.

That is where our unfortunate new friend agenticness enters the room.

It describes, roughly, the degree to which an AI system can pursue a goal through actions rather than merely responding to instructions one step at a time. And there can be more or less of it.

But there is another word nearby that matters even more: agency.

If I say, “Publish this,” the intention is mine. I decided that this thought mattered. I decided that it should become a piece of writing. I read it and approved it. I decided where I wanted it to go.

Molly may then have considerable freedom in accomplishing what I asked her to do. But she didn’t wake up that morning desperately wanting to publish my writing.

That distinction seems increasingly important.

Perhaps the interesting future isn’t one in which we give machines as much agency as possible. Perhaps it is one in which we become increasingly sophisticated about sharing agency.

I provide intention, curiosity, judgment, values and permission. Molly provides memory, research, organization, technical ability and an increasingly remarkable capacity to act.

And meanwhile, the interface between us becomes strangely simple.

We talk.

Which is another reason you should probably get used to hearing agenticness.

Sorry.