What Do You Mean by “Done”?

In the early stages of working with a new client, I didn’t start by defining terms.

I started by listening.

People described their processes, goals, and systems using words that sounded familiar: “approval,” “done,” “process,” “complete.”

At first glance, everything made sense.

But I had seen those same words used very differently in other organizations. Sometimes “approval” meant a formal sign-off. Sometimes it meant a quick nod. Sometimes it meant nothing at all.

So I listened for patterns, then stepped in.

“When you say ‘done,’ what has to be true?”

“When you say ‘process,’ is that written down?”

“What actually happens if someone skips this step?”

Those questions weren’t about semantics. They were about stability.

Most problems didn’t come from bad decisions.

They came from people using the same words to mean different things—and not realizing it.

There’s a reason for that.

Concrete things are easy to agree on. A wheelbarrow is a wheelbarrow. You can point to it. If something is missing, most people will notice.

Abstract terms are different.

Words like “done,” “approval,” “quality,” or “complete” don’t have fixed boundaries. They depend on context, expectations, and experience. Two people can use the same word and mean something slightly—or significantly—different.

But in conversation, those differences are easy to miss.

So we talk about abstractions as if they were concrete objects.

We argue about them as if they had clear edges.

Most of the time, this goes unnoticed—until something breaks.

A task marked “done” isn’t actually finished.

An “approved” step gets revisited.

A “complete” process turns out to have gaps.

The result is confusion, rework, or disagreement about what should have happened.

Not because anyone intended it.

But because the meaning was never fully aligned.

Before any system, process, or decision can be reliable, the meaning behind its key terms has to be stable.

That doesn’t mean perfect agreement.

It means differences are visible.

Assumptions are surfaced.

When someone says “done,” there’s a shared understanding of what it includes.

This isn’t just a communication issue.

It’s structural.

When meaning isn’t stable, everything built on top of it becomes harder to trust.

This problem existed long before AI.

But AI makes it easier to miss—and harder to correct.

Subtle Link Paragraph

If this feels familiar, it shows up in a different way when working with AI systems—where the language stays fluent, but the underlying meaning can shift over time. That’s part of what I explore in Not Smarter, More Stable.