AI versus Marshy #51: counterpoints, that voice survey, and building an AI business from scratch

Hello and welcome to the next edition of AI versus Marshy!

This is the newsletter that stays in touch with the AI hype machine, and boils things down to: what’s actually happening? and, what’s useful to know right now?

The dust has settled on our 1st birthday, and it’s back to regular business.

This week we cover:

  • A counterpoint to the AI hype machine
  • So what happened with the VoicePanel link?
  • Building an AI business from scratch - Update #1

Lots to get on with this week so let’s make like a pram and roll on with it.


A counterpoint to the AI hype machine

Via Internet of Bugs.

Meet Carl.

Unassuming - I like it

Carl is a software developer with 35 years experience that popped into my YouTube algorithm the other day.

My algorithm is fairly (I quote my beloved Georgie here) - “quite nerdy” - as you can imagine.

Carl’s video called AI Hype is completely out of control - especially since ChatGPT-4o is what caught my attention and it was refreshing.

The video is 22mins of detailed nerdy proselytising, so let me break down for you what’s interesting about it instead.

Carl is coming at the AI wave through the lens of a software developer. He’s employing cold, rational data to his opining - starting with research and peer-reviewed papers, and then cycling down to less robust sources of information. All information is fallible over time but he’s leading with evidence

Carl dismisses a lot of the hype about “human-level artificial intelligence” as nonsense and highlights that nearly all of the tech demos we’ve seen are staged

Carl and I agree that nobody knows what’s going to happen, and what to look out for:

What’s going to happen with AI is one of the bigger questions and software careers right now. In tech. Maybe even in the world, and I’d really love to know what’s going to happen. And if I did, I would tell you, but I don’t. And honestly, nobody does.


And what we really know for definite sure is that many of the people that are telling us how great it’s going to be, not only have a financial incentive to lie about that, they’ve been lying about it for years and have been lying so obviously about it that they’ve been caught lying about it over and over.

There’s some other interesting points about ChatGPT 4o being worse than 4 in some tests, and many cited examples where all the big tech companies have lied.

I think where I differ is coming from the optimist POV - I believe the gap we’ve got in the world today isn’t with capabilities it’s with tech literacy and its application.

Despite massive increases in STEM education - tech is still not fully appreciated/comprehended by the broader world.

There’s entire industries and geographies missing out, and if more time/energy/effort can be directed to addressing the gap - I think that’s where we win (instead of more breakthroughs).

Want to know the other interesting part about our dry delivering developer schtick?

Carl has released 20 videos, the first one 3 months ago, and now has 45k subscribers, and 1.2m+ views.

In the world of SEO its accepted that it can take 6-12 months to truly start seeing traction.

He’s gone hard at a niche he is passionate about with YouTube and has built a platform in record time.

What I love about it is it’s not a crazy production value masterpiece (if you want to know what extreme in this world looks like watch a recent MrBeast video).

There will always be markets for passion and creativity.

What Voicepanel does (or why Marshy sent a survey link earlier in the week)

I first mentioned Voicepanel in edition 38.

For those of you who couldn’t access the survey while it was open - it’s a tool that gathers survey feedback via voice.

This is useful because:

  • It lets people answer in their own style and voice
  • It’s faster than most in-depth surveys
  • There’s richer answers, as its AI can drill-down for specificity (which is something a good interviewer will do too)

Nothing replaces 1:1 interviews - but this gets closer to that benefit than you ordinarily would from a straight up survey.

Survey data = quantitative

Interview = qualitative

This approach is almost like a blend of both as you’re uncovering themes and nuance.

I was testing Voicepanel for a new thing (see below) but wanted to share what it looked like on the back-end for my readers.

Before I do though - a big beautiful THANK YOU to those of you who answered, it’s sometimes a lonely job writing a newsletter as it’s much less interactive than other writing (say - a social media post or a Slack community).

This isn’t restricted to small newsletters like mine - I wrote to my friend Bill about a great example in his newsletter about building systems and was humbled by his similar response:

This wasn’t emotional bait for replying to this email (I got a great dose earlier this week) - but if there’s another one you’re enjoying be sure to reply and let the author know.

This tip also applies to non-fiction book writers.

So what’s the back-end of Voicepanel look like?


I got to see how many visits to the link were made.

Then I got to see the response rate and how long it took people to answer (thanks 11minuter!)

I can go deep on answers and see the responses, as well getting a summary. Importantly - I’m not getting email IDs or identifying information. This is an option for higher-tier versions of the product, but unnecessary for what I’m building below (see next article).

Finally - if going through responses is too large a job - I can also get a summary report via the Analyze tab.

Why is this relevant for marketers and researchers?

I’ll give you an example with a service provider marketing themselves.

There’s always a gap between what a service provider says about themselves versus what their clients say about them.

This usually is a non-issue - once someone becomes a client, they have a much deeper relationship with their service provider that isn’t reliant on just what’s being said.

When we’re marketing though - we’re looking to attract new clients (people who don’t know the service provider).

The gap between what the service provider says and what the clients say becomes far more important.

This is because we want to understand what led the client to work with the service provider before they knew them.

A tool like Voicepanel can be designed to capture this information, which is a good bridge into my next AI project.

Building an AI product from scratch - update 1

One of the things that fascinates me about AI is the ability to do more with less.

We’ve seen this applied so far with varying degrees of efficacy.

Some skills (like transcription) are orders of magnitude better than if they were done by a human.

Other skills like authentic content creation - aren’t in trouble any time soon.

I have a family. I want to work for myself and support them. I do this already but don’t want to be at the mercy of referrals and custom consulting work.

These are all very good reasons for

It’s a service model I’m building with scale in mind, predictability, and running on my own.

I think running something like this solo was very hard 5 years ago, and I think it’s more possible today.

I’ve soft-launched it this week, and am putting it through its paces.

It’s built with Carrd, Beehiiv, Airtable, Google Appointment Scheduling, and is using Voicepanel and Design Pickle (planned) for fulfilment.

I coach a lot of startups - one of the things I coach them on is not running ads because they don’t know what they’re doing. I’m breaking my own rules here because I do know how to run ads, but also understand this service, page, and offer isn’t validated.

I’m going to run ads through this over time with the goal of achieving predictable conversion.

At the start my numbers will be pathetic and disappointing.

I also know I can shift many things to refine this: my service, my offer, my process, my ads, my copy, my pricing, and much more.

I’m planning to cycle through this process in 1 week sprints. Fully trusting that will be in much better shape by weeks 10, 50, and 100.

I’m planning to #buildinpublic and will update the numbers each week via my LinkedIn.

I also intend to continue updating things in here from an AI perspective.

My stack is very deliberately built with automation and AI in mind.

For example - I created a private “concept creator” GPT that’s been fed knowledge I’ve acquired over the years about lead magnets.

I can use this to get started on orders that come through.

GPTs are still funny to use but one of the improvements is getting them to “take actions” outside of the interface.

What I will be able to do with more time and information coming through is automate form submissions (currently located here) to connect into my GPT and produce output that moves the creation process forward.

Not only this, but I can keep providing feedback and training in plain words about what it’s producing. So it will get better over time.

This is without any specialist coding knowledge (to be clear I don’t have any).

My ultimate goal is to build this myself so I can show others to do it as the capabilities improve over time.

I know I’m not the only one who wants to work for themselves 😉

That’s it for this week!

On the personal front our twin boys (and in turn ourselves) have been battling waves of daycare sick.

It’s not fun but I know the boys are enjoying the extra stimulation from going to daycare and playing with other toddlers.

Was it a good issue? Let’s keep going!

-Marshy 💪

AI vs. Marshy

Growth marketer meets biggest technological advancement in our lives. Learn about AI in a way that doesn't overwhelm. Add a splash of strap yourself in and be prepared.

Read more from AI vs. Marshy

Hello and welcome to another edition of AI versus Marshy! This is the newsletter that demystifies the AI hype and gives you actionable and practical information instead. This week we look at: • Researching on autopilot with Zapier Central • Consultants an early winner in the AI race • Sprint 4 update - Now there's lots to crack on with today, so let's make like a whip and flick to it. ChatGPT drawing a whip having a crack -Marshy Research on autopilot with Zapier Central...

Hello and welcome to another edition of AI versus Marshy. This newsletter tracks what’s happening in AI and applies a practical filter for our readers. But don’t take my word for it - here’s what one of our readers said: “It keeps me up to date with what’s going on in AI, without having to do the heavy searching myself.” This week we cover: Using AI to reverse engineer content trends (h/t Starter Story) Figma is coming after Powerpoint - with some backpedalling Sprint 3 of

Hello and welcome to another edition of AI versus Marshy 👋 This is the newsletter that is excited about AI, but keeps it grounded in things you need to know and can do today. This week we look at: • Using ChatGPT for fashion advice • Editing pictures with words • Another build in public update #2 - Lots to share with you today, so open like a door and make room. -Marshy ChatGPT for Your Own Wardrobe Via Ruud Hendriks on LinkedIn. My friend Anna shared this practical example...