The short version: I record myself talking, usually on a walk with my dog, then run the transcript through AI to turn one ramble into a newsletter, three LinkedIn posts and a handful of email replies. It works because talking is easier than writing, not because the AI is clever. The AI is just the typist. You still have to have something to say.
The blank page problem nobody admits to
Here is what nobody tells you about working from home and running your own small business: the hardest part isn't the client work. It's sitting down to write the newsletter, the LinkedIn post, the email to your list, when you've already spent your brain on calls all day.
I used to block out three hours on a Sunday to write for the week. Three hours, most weeks, produced one decent blog post and a lot of guilt about the ones I didn't write. I've written before about what happened when I tried to solve this with sheer volume, publishing 569 posts in 45 days and watching Google index about 5 percent of them. That taught me volume without a voice behind it is worthless. So the answer isn't write more. It's write differently.
The answer I landed on, about eight months into rebuilding my business publicly, was to stop typing first drafts altogether.
The actual routine
Every weekday morning, roughly 7.30am UK time (earlier if I'm in Israel, where I spend a good chunk of the year), I take my dog Milo out for 12 to 15 minutes. I press record on the Voice Memos app on my phone and I talk. Not a script. Not notes. Just whatever is on my mind about the business that day: a client problem I solved, something a prospect said on a call that annoyed me, a thought about a tool I tried and hated.
Here's the step by step, exactly as I do it:
- Step 1: Pick one thing before you press record. Not "content ideas", one specific thing. "Why the AI chatbot my client installed lost them a sale on Tuesday" is a topic. "AI and customer service" is not.
- Step 2: Talk for 10 to 15 minutes as if you're explaining it to a friend who runs a similar business. Say the numbers out loud. Say the client's actual objection, not a paraphrased version.
- Step 3: The moment you're back, drop the file into a transcription tool. I use Otter.ai, which is about £8 a month on the basic paid plan. A 12-minute recording comes out as roughly 1,800 to 2,200 words of raw, messy, repetitive transcript.
- Step 4: Paste that transcript into ChatGPT with a prompt that says something like: "This is a transcript of me talking, not a finished piece. Keep my voice, my opinions, and my specific examples. Cut the repetition and the filler words. Do not add new claims or invent numbers. Turn this into a 700 word blog post in first person." That last instruction matters more than any other part of the prompt.
- Step 5: Read it back and change anything that sounds like it came from a robot rather than from me. There is always something. Usually it's a sentence that's too tidy, too balanced, too "on one hand, on the other hand" for how I talk.
That's the whole system. No fancy software, no course, no six-step funnel. A 12-minute walk becomes a blog post, and the same transcript usually gives me three LinkedIn posts and one or two lines I can use in a client email, because I already said the thing out loud once, which is most of the work.
The morning it worked
I'll give you the specific one, because vague examples are useless. In March I had a call with a client, a woman running a 14-person recruitment firm in Manchester, who'd spent £4,000 on an AI-powered chatbot for her website. The chatbot answered a candidate's question about salary bands so badly and so robotically that the candidate messaged the recruiter directly to complain, and the recruiter nearly lost that placement. My client was furious and, frankly, so was I when she told me.
On my walk the next morning I talked for eleven minutes about exactly that: the £4,000, the specific bad chatbot answer, why she'd bought it in the first place (a sales rep told her it would "save her team 10 hours a week"), and what she should have asked before signing the contract. I said the number, I said the industry, I said what went wrong, in that order, without planning it.
By 9am that transcript had become a 900-word blog post, two LinkedIn posts (one about vetting AI vendors, one about the actual chatbot script problem), and a paragraph I lifted almost word for word into a follow-up email to a different prospect who was asking about the same thing. Four pieces of content from one walk. The blog post ended up being one of the pieces that helped bring my organic traffic back after two flat years, which I've written about in more detail elsewhere on this site.
Total time from dog walk to four published pieces: about 55 minutes, including my coffee.
Where this goes wrong for most people
I've watched other business owners try this and give up within a week, and it's almost always the same three mistakes.
First, they ramble without a target. They press record and talk about "the business" generally for 20 minutes, and the transcript is unusable because there's no single point in it. AI can tidy up messy sentences. It cannot invent a point you never made.
Second, they let the AI output go out unedited. This is the uncomfortable bit that most people rushing to sell you "AI content in minutes" don't want to say out loud: if you publish the first draft the AI hands back, it reads like everyone else's AI content, because it is trained on the same patterns as everyone else's AI content. The editing pass, the bit where you put back your specific opinion or your bad mood about that chatbot, is not optional. It's the whole job now. Recording and prompting is maybe 20 percent of the work. Editing your own voice back in is the other 80 percent.
Third, and this one stings: they think the tool fixes a business that has nothing to say. If you don't have opinions, client stories, or numbers worth mentioning, no amount of talking into your phone will produce good content. The method makes an interesting business faster to write about. It does not make a boring business interesting.
What this costs and what it saves
Real numbers, because vague ones are what make this stuff hard to trust. Otter.ai's paid plan runs about £8 to £10 a month depending on currency and plan tier. ChatGPT Plus is $20 a month, or you can do the same thing with a free tool if you're patient with the limits. That's under £30 a month total.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Against that: I was spending roughly 10 to 12 hours a week on content before this, across blog posts, LinkedIn, and email. I now spend around 3 to 4 hours a week, most of it editing rather than staring at a blank document. If your time is worth even £50 an hour, that's a saving of somewhere between £300 and £400 a week, for a £30 monthly outlay. I don't say that to sound like a sales page. I say it because when I tell other consultants this, their first question is always "but is it worth it", and the honest maths says yes, easily, provided you do the walk.
Where the transcripts go beyond blog posts
Once you have a habit of daily voice notes, the transcripts become raw material for things people don't think of as "content" at all. I've pulled paragraphs from old walk transcripts into guest post pitches, because an editor wants a specific, opinionated angle, not a generic one, and my rambling morning thoughts are usually more specific than anything I'd sit down and plan. If you're trying to get a guest post accepted, a genuine, slightly rough opinion pulled from a real conversation beats a polished pitch every time, which is a point I go into in my guide on getting guest posts accepted.
The transcripts have also become the source for FAQ sections on my own site, because when I talk out loud I naturally answer the questions people ask me on calls, in the order they'd ask them, rather than the tidier order I'd invent if I sat down to write a list of FAQs cold.
If you're not confident with any of the tools involved, that's fine, you don't need to be technical to run this system. I wrote a full walkthrough of how I use AI day to day without being a techy person at all, which covers the basics if Otter or ChatGPT feel intimidating right now.
If you'd rather someone build this for you
I get asked, more than I expected, whether I'll just set this system up for a business owner rather than explain it. Some people would rather pay for a working system with their own prompts and their own transcription workflow than spend a month tinkering. If that's you, it's the kind of thing an AI implementation coach would build with you in an afternoon rather than the six weeks it took me to get it right on my own through trial and error.
Either way, the tool part is the small part. Otter or Whisper, ChatGPT or Claude, it barely matters which. What matters is whether you say something specific enough on the walk to be worth typing up at all.
A few practical extras that make the difference
- Record in a quiet-ish spot for the first two minutes at least. Transcription tools mishear words badly against wind and traffic, and you'll spend more time fixing "chatbot" turning into "chat bought" than you saved.
- Say numbers and names out loud, deliberately, even if it feels odd. "Four thousand pounds" transcribes and reads better in a draft than a vague "quite a lot of money" that you'd have to go back and fix anyway.
- Keep one recording per topic. Combining three ideas into one 20-minute ramble produces a transcript that reads like three separate, half-finished articles glued together.
- Don't skip the edit. If you're short on time, better to publish one edited piece a week than five unedited ones nobody can tell apart from the last AI tool everyone else is using.
I still keep a running note of ideas I stumble across from other people, the way I found half my current tools from a list of odd, useful websites I'd never heard of before someone shared it with me. Half the battle with any new habit is just having the right small tools sitting ready before you need them, so you're not hunting for a transcription app at 8am with a dog straining on the lead.
I have written more around this on the site: [Trending Topic 2020] How to engage and convert more visitors with interactive calculators, I Published 569 Blog Posts in 45 Days. Google Showed 5 Percent of Them. Here Is the Honest Math on AI Content., The Month My Website Started Paying Me Again.
Frequently asked questions
Do I need expensive software to do this?
No. A phone's built-in voice recorder, a transcription app costing under £10 a month, and a free or low-cost AI chatbot is all the system needs. The expensive part isn't the tools, it's the time you'll spend the first few weeks learning to talk to your phone without feeling odd about it.
Will AI-written content from a voice memo sound like me?
Only if you edit it. The first AI draft will sound generic because it's trained on generic patterns. Your job is to put back the specific opinions, the exact numbers, and the slightly blunt bits that sound like you rather than like a press release.
How long before this saves me time?
Most people see the time saving within two to three weeks, once they stop overthinking what to record and just talk about whatever annoyed or interested them that day. The recording part gets faster with practice. The editing part stays roughly the same length, which is the part worth protecting.
What if I don't have anything interesting to say on a given day?
Then don't record that day. This method amplifies a business that has real client stories, real numbers, and real opinions. It does nothing for a business that has none of those, and no amount of clever prompting fixes that gap.
Related reading: AI Notetakers Are Quietly Costing You Client Trust. Here's What I Do Instead and What Is ChatGPT Telling People About Your Business Right Now? (Probably the Wrong Thing).
