By Anna Woolliscroft, Content Writer & Founder, Market Avenue Ltd
You probably fall into one of three AI camps right now.
Camp one: you’re pretending it isn’t happening. Camp two: you’ve dabbled by asking it to tidy up a LinkedIn post or help you plan dinner from the fridge leftovers. Camp three: you’ve got a subscription, you’re using it regularly, and you know there’s more. You’re just not sure what more looks like in practice.
If you host a podcast, this article is for you. The same thinking applies to any content you produce repeatedly, such as proposals, newsletters, social posts. But let’s start with the podcast.
I’m a communications specialist, and by my own admission, a perfectionist who over-thinks. It’s a personality type that means tasks take longer than they should, and nothing ever feels finished. Thankfully, AI has become the thing that gets me out of my own way.
Here’s an honest case study
Each episode of my Writing with Purpose podcast used to cost me five to six hours of editing and post-production content work. I listen back to every episode in full and repurpose the content across multiple platforms. Quality matters to me, so I was never going to cut corners, but the hours kept stacking.
So I built a custom AI podcast assistant. I gave it a role and context about the show, uploaded my tone-of-voice document and a language style guide, then trained it on exactly what I’d hand over after recording and what I needed back. The post-production workflow now delivers show notes, a blog post, promotional clips, social posts for four platforms, a YouTube description, an email announcement, quotable content, and a guest follow-up message.
‘Your promo materials are the best I’ve seen from any of the podcasts I’ve been involved in… it gives you insights you wouldn’t ordinarily have.’ — Nic Wilson, writer, editor, Guardian country diarist
The assistant took around a day to build. After working on four episodes, each one now takes two hours less to complete. Producing two episodes a month, I’ll save 48 hours a year minimum. It paid for itself in month one.
‘Good is good enough,’ and this framework gets me there without the overthinking.
Five steps to build your own assistant
1. Identify one repeating content task that currently costs you time.
2. Get a paid subscription to an LLM (ChatGPT, Claude, Gemini).
3. Write a tone of voice document. Even a rough one. What do you sound like? What would you never say?
4. Build your assistant around one specific workflow. Give it a role, context, and a clear output. Test and refine.
5. Build in human review checkpoints to stay in control. This is not optional.
I’ve identified ten AI assistant projects across my work. If each saves a conservative 40 hours a year, that’s 400 hours annually, and over ten working weeks back in the calendar.
Set up once. After that, it just works.
This article was written with AI assistanc>
