Verification Bandwidth

The other day I talked about taste being the edge in the AI era

That as execution gets cheap, the human job becomes conducting, and the successful coaches will be the ones with the taste to judge what comes back.

This is the next question along - how do you actually verify all that work?

Because if AI keeps going the way it’s going, you won’t have one assistant - you’ll have dozens of pieces of work running at once.

Programmes being drafted, check-ins being summarised, content being cut, emails being written.

And I use those examples as if the work that we do is going to stay the same because they’re things that we do now.

In reality, I think it’s possible that we’re going to have a lot more data and inputs as more clients use wearables.

The wearables give us more information, and they want us to combine our expertise with AI to use that data to inform what we do moving forward.

The standard worry is attention.

How does one human possibly keep up with all that output?

But I don’t think attention is the constraint.

Humans solved that problem centuries ago - no CEO reads every email in their company.

Delegation, hierarchy, only the exceptions reaching the top.

Here’s the thing: the real constraint is verification.

Work scales with AI at exactly the speed you can check it.

Coding is exploding right now because checking is cheap - you run the code, it works or it doesn’t.

Coaching isn’t like that.

What does ‘done’ look like for a client message?

A programme? A piece of content?

There’s no test to run, we have to read it and judge it ourselves.

Which means the coaches who get the most out of what’s coming won’t be the ones who can generate the most - that race has already ended, generation is almost free for everyone.

Speed of verification is going to be key - knowing your standard so precisely, that you can look at a piece of work and call it in seconds.

Taste tells you what good looks like.

Verification bandwidth is how fast you can apply it, over and over, without it degrading.

It’s a different skill, and I’m not sure anyone is even thinking about it.

So, the prediction - within a few years, the gap between a coach running one AI and a coach running fifty won’t be technical skill or budget.

It’ll be verification bandwidth - how much work can pass through your judgment in a day before the quality starts to slip.