We’ve been AI sceptics, AI critics and AI side-eyers here at WFB – we’ve kind of got a vested interest in keeping good writing around. But we’re also pragmatic at heart, and we’re always looking for tools to help us write better, hit the mark on client briefs, and cut out some of the fiddly, annoying jobs.
Let’s be clear: we don’t use generative AI (GenAI) to do our writing for us (never!) Or even to draft things. We do use it for things like transcribing and summarising interviews, pulling out details from super long documents, checking our work, and formatting docs. Precise, detailed jobs that take a long time for our squishy human brains to do.
Here’s what that looks like:
Interviews are a big part of our job. Even just two years ago, we recorded interviews and then transcribed them ourselves – hours of just furious typing. The transcription software around at that time just… wasn’t good. Transcripts showed up as huge blocks of text, with no way to tell who was talking. They weren’t accurate (especially for Kiwi accents). This meant lots of back and forth between recording and transcript, and a lot of time spent pulling out quotes and details before you could even start writing.
Now, it’s hard to remember those bad old days. We use Otter AI, a creepy little bot that joins and records all of our video calls as they happen (you can also use it for IRL meetings). It’s cut literally hours of tedious work – we’re never giving this one up.
No more impenetrable blocks of text, more:
Some of our clients are a tiny bit finicky about their content. They have rules about hyphens, capital letters, how to spell certain words, and whether to use an Oxford comma. Some are based in the US, which means fewer U’s and more Z’s. Most have a specific tone of voice that they’ve developed to fit their brand. Important? Yes. Easy to remember? Not so much.
We used to keep lists or create Wikis to help us keep track of client TOV rules and guidelines, but that meant a lot of clicking around between docs and remembering to update them. It’s so important and so easy to mess up that we’ve developed our own tool to help.
We create voice and style guides for our clients and add them to WFB Voicey. Then each document gets checked against those rules.
It pulls us up on:
Before we can even type a headline, we’re often wading through a folder of interviews, background info, past client work and the brief itself. All that information is critical, but our squishmallow human brains aren’t designed to consume it all at once. That makes it too easy to miss details or context.
So, we’ve just started using Google LM Notebook. Once we’ve read through everything we load all the docs into a project. Having all briefing documents in one place is a good thing, but it’s not what makes this AI tool so useful. That would be the chat feature, which answers your questions by pulling information from the documents provided. Crucially, it adds citation markers, so you can see where it’s got the info from. Want to know what an interviewee said about a certain topic or which services the client provides? It grabs answers only from your docs, so you know they’re relevant.
Useful for:
Don’t get us wrong, we’re still all over Google Scholar, but Perplexity is a really good addition to our researching processes. It prioritises accuracy, surfacing research and references from across the net. Unlike ChatGPT, it uses up-to-date info, is less likely to invent things if it can’t find what you’re looking for, and, perhaps most crucially, isn’t a suck up.
Delivers us:
Lovable (and tools like it) will spin up little bits of software based on your plain-language instructions. It’s frustrating to use (although that’s probably because I’m a coding novice), and the tools are basic and buggy. Still, in a couple of cases, it’s been totally worth it. For example, it’s breathtaking how long it takes us to pull copy off a website and format it manually. We need those Word Doc versions, so we can accurately price up a project and when we’re tweaking existing copy. So, I spent an hour or so knocking out a little tool to do it – it grabs the text a user sees, ignores footers and system text, and uses OCR to format it in a way that approximates the hierarchy and layout of each page. Basic stuff, but revolutionary for us.
Epic at:
If you think about it, typing is a pretty inefficient way to transposing your thoughts into writing. We’re fast, but still the fingers lag way behind the brain. The voice is a little better, so in many cases, using Whispr Flow is a gamechanger. It’s dictation software installed on your laptop and mobile. Speak aloud and it’ll perfectly transcribe what you say with accurate spelling and punctuation. It can even handle the Kiwi accent and words in Te Reo Māori. Smart. The volume of words you can get down is incredible, but that has a downside too – if you’re going to send that rambling text or Slack message, you might as well make it a phone call. I find it really useful when I’m bossing around an AI tool or knocking out a first draft. It’s flipped my process for writing dialogue too. Instead of testing my script by reading it aloud, I start with natural speech, then tweak out all the clumsy bits.
Use it for:
We’ve been saying the same thing for ages – AI is a useful tool, but it’s not really intelligent. That’s why we use it for the nitpicky details rather than the zingy sentences, thoughtful insights or heartwarming stories. Think getting a robot to do your laundry, not play with your kids.
The upshot: We’re getting the grunt work done more quickly, which gives us more time and brain space for writing. And we really need that brain space.
The lead copywriter and creative director at Words for Breakfast. She spends much of her time working with the copy writing team across a variety of projects.
