I’m sitting in my backyard on a lovely, late summer afternoon. Soon, the marching band will begin their fall practices on the high school football field just down the street, and parents and students will be driving past my house in droves. So I’m enjoying these last few days of quiet, before all the busy-ness of fall begins.

For some strange reason, fall is always the busiest time for me for freelance work, too. Why the heck does everyone want a podcast that launches in the fall? Can’t some of you people launch in the spring?

And this fall, I am SO BUSY.

I feel awkward admitting to that right now. I know that there’s a lot of unemployment and upheaval in this industry. And, to be perfectly transparent, the first six months of this year were very slow for me. But for the last couple of months, I’ve had the opposite problem. Too much work to do, and not nearly enough time to do it.

And that’s okay. The truth is, if our little narrative industry is going to survive the current upheaval, we’re going to have to figure out how to do more work with less time.

And note, I’m not actually saying that we need to do “more work for less money.”

Yes, less time = less money. But too often in our industry people think that saving money means, “Hire qualified people and pay them less per hour.”

And that, my friends, is not the answer.

But we do need to do something to balance the books! And look, narrative podcasting doesn’t have an audience problem. There are more listeners every year. And it doesn’t have an advertising problem. Advertisers exist. They’re willing to spend money.

The problem is that the value of the advertising (and subscriptions) often doesn’t cover the cost of production.

This is where the “less time” idea comes in. Because if I can produce two podcast episodes in the amount of time it used to take me to produce one, everybody wins.

And I really do mean everyone. This isn’t just a way to make sure that I stay employed while other podcast producers find new industries to work in.

There is no limit to the number of people, companies and organizations that WANT their own podcast. And if we can find ways to make it more affordable – and maybe even, gasp!, profitable – everyone who works in this industry can have healthier, happier bank accounts.

I’ve already written a long post about all the ways that AI will never replace me. And I still firmly believe all of those things to be true.

But that doesn’t mean that I don’t use AI. There are lots of ways that I use AI to make my job more efficient. Easier. Faster. Less expensive.

And let’s be clear, these tools aren’t replacing the parts of my job that I love.

They are replacing the parts of my job that I loathe. Truly loathe.

I’ve already talked about all the ways I use Descript to speed up my production process. But this time, I thought I’d run through some of the other ways I’ve been using AI to make my work more efficient.

And don’t worry…no clients have been shortchanged by these innovations. I charge by the hour. So if I save time, they save money!

AI Transcriptions

Let’s start with the easiest and least controversial. Audio producers have been using AI powered transcriptions since long before AI became a buzzword. But I STILL see scripts where the audio has clearly been transcribed by hand.

If this is you, stop. Just stop.

In the early days, I used to tell people that machine transcription made too many mistakes and insist that “purists” log their tape by hand, in order to become deeply familiar with their tape.

But…let’s face it, that’s a bunch of bullshit.

Because you’re not really becoming “deeply familiar” with your audio when you have to pause every few seconds to let your fingers catch up. You’re just not!

Instead, I let AI transcribe my audio. And then, I listen through to every single word, at normal speed, highlighting the parts that I think I might want to use.

It’s much more efficient, and just as effective.

AI Headlines and Episode Titles

I am good at many things, but I suck at writing headlines.

Suck, suck, suck.

And episode titles? Sometimes I can come up with something brilliant. But usually, my brain goes blank as I stare at that little flashing cursor on my screen.

Do you ever get the sense that the cursor is mocking you? I swear, I hate that thing sometimes.

Now, I’m not going to say that AI is brilliant at writing episode titles.

AI has no taste. It’s not discerning. It will spit out episode titles that absolutely, positively, should not be used!

But AI is really, really good at breaking through writer’s block. Show it your script and ask it for 10 possible episode titles. Then, take your favorites and tweak them until you hit upon something you love. Easy, peasy.

Take that, flashing cursor!

AI Episode Descriptions

I am a “show, don’t tell” tyep of writer. I am really good at building a series of scenes that SHOW you why an episode is important and what you should take away from it.

But when I have to just sit down and TELL you why you should listen, I am at a loss. All of my “narrative first” training works against me, and I’m stuck.

Suddenly, I’m not sure my episode is interesting or important anymore. I don’t know how to inspire people to listen. And I sit there, staring at that stupid flashing cursor again.

Enter AI.

When I’m running my script through AI, asking for ideas for a headline, I also ask for a “short description” and/or “show notes.”

The version that AI gives back to me feels like the first draft of something that’s been written by an intern. It’s not right, yet. But it’s a start.

And a start is all that I need to keep that annoying flashing cursor at bay.

AI Research

Okay, I’m not gonna lie. This is simultaneously the most useful way to use AI – and the most dangerous.

Let’s start with the danger. Not so long ago, I was researching a story that involved medical data. And, not surprisingly, I was having trouble finding actual cases, because they’re all confidential under HIPAA.

(For the non-American’s among us, HIPAA is a US law that forbids medical professionals from disclosing information about their patients. You probably have one of these laws where you are, too.)

So, anyway, I asked AI to provide me with a list of cases, and it spat out a list of 7-8 examples. They were all relatively detailed, but lacked any identifiable information, like names, dates or exact diagnosis. Still, if they were real, they would have given me a big head start in my research.

Eureka! It’s a jackpot!

But then I asked AI to cite its sources. And instead, I got a civics lesson about HIPAA and the fact that medical data is not public.

Uggghhh.

It turns out, AI completely made up all those cases. Out of thin air. They did not exist. And that was more than enough to make me skeptical about AI as a research tool.

But recently, I decided to try again.

I’ve been putting together an interview slate for an existing podcast. And as I’ve spoken about many times before, I like to think about ALL the diversities when putting together a slate of episodes. In this case, I was looking at all of the episode ideas on my list and I realized I needed something:

• About Subject A.
• In Region B.
• With a woman as the central character.

So, I asked ChatGPT for a list of women who had been involved with Subject A in Region B.

I got a short list of people. So, for each of them, I asked.

Describe the work Woman #1 did in the area of Subject A in Region B. And provide your sources.

And wow…we hit pay dirt. The third woman on the list had been deeply involved in a tale so twisty, it made my inner storyteller sing.

But I’ve been burned before. So I went directly to those sources and tried to search for the woman’s name, to confirm that AI wasn’t just making stuff up again. And lo and behold…

No results.

Seriously, there was not one single source that confirmed the work Woman #3 did in Subject A in Region B.

So, me being a smarty pants, I wanted to let AI know that it was wrong ... again. Politely, of course. So I went back in with another question.

I would like help finding specific reports or press statements with Woman #3’s name, especially ones that reference her work in Subject A in Region B.

And I got this response:

I couldn’t find any reports or press statements that specifically cite Woman #3 by name in connection with Subject A in Region B. However, here’s a list of sources that note actions taken by “Job Title,” which, during the years of the incident, would refer to Woman #3.

Shit. Seriously? I never would have thought to search by her job title. Never!

And yes, I verified the details. And yes, AI was correct. (This time.)

And so, not only do I have a great episode for my podcast, I now have a new way to think about how I search for information.

Win-win!

So that's how I've been using AI lately. I’d love to hear from you. How are you using AI to improve your workflow? What are your hopes for it? What are your fears? And where do you draw the line?

Our next Narrative Beat Monthly-ish Chat is scheduled for Wednesday, September 3 at 1pm Eastern. Remember, these are totally free. I just ask that you RSVP.

Hope to see you there!

And as always, please share this newsletter with anyone you know who needs it. It always makes me happy when our little community grows!

Karen