Episode 11: AI Makes Bad Direction Cheaper (with Maury Postal)

It probably goes without saying that I rely pretty heavily on AI these days.

I do a lot of writing and publishing and I have zero audio visual editing skills of any kind.

So being able to use AI tools to help me edit a podcast is genuinely useful.

And at one point, I thought I had completely nailed the process with Claude.

We were working through these increasingly elaborate editing guides.

They were color coded.

There were indicators for what to cut, what to add, what to record, what needed a light patch, blah, blah, blah, blah, blah.

They were complicated, but they worked.

And I want to note, they weren't complicated because I made them complicated.

Every time we did another round, Claude would add some kind of new graphic, new feature, new indicator.

And I was like, oh, this is helpful.

This is easy to read.

I save one.

I tell Claude, this is the template.

We're using this going forward.

Easy breezy, except apparently asking Claude to use the template that Claude created was becoming increasingly challenging for Claude.

Every time I uploaded a new transcript, I got a different system.

I had new colors, new categories, new terminology.

And Most frustratingly, words would just disappear from the transcript, which was a problem because preserving every word and telling me what to do with it was literally the entire point.

I needed an accurate transcript that showed me what to cut and fix so I could make those changes and descript my editing software.

And what I got was a beautiful, organized, increasingly colorful document that did not do the job I needed it to do.

But don't worry, it used way more of my session usage or memory.

So it also slowed Clawed down, which is not helpful.

And The output looked better, the process was questionably faster, but the underlying material was wrong.

And I keep thinking about that because it kind of captures one of the stranger risks of working with AI.

It can produce something polished enough that you don't immediately notice it has solved a completely different problem from the one you actually had.

Two weeks ago, my friend and former colleague, Maury Postel, joined me to talk about why the creative brief isn't paperwork before the real work.

The brief itself is the work.

But our conversation kept going because Maury and I have never been especially gifted at ending a conversation.

expediently.

And once you start talking about direction, alignment, and process in 2026, you inevitably end up talking about AI.

So this is part two.

You don't need to have heard part one first, although obviously I would love it if you did.

This time we're talking about what happens when AI makes execution faster and cheaper, but doesn't necessarily make the direction any better.

I'm Karen Doak, and this is OK, Actually, a show about what's really going on at work and what we can do about it.