How I built an AI version of Scale at Speed
For the past twelve years I’ve been running a well-regarded accelerator programme for service businesses, mainly marketing, digital, creative and comms agencies.
In 2021 I wrote a book called Scale at Speed – How to triple the size of your business and build a superstar team. It was published by Hachette+Robinson and a lovely man called Roger Davis did an audiobook version. Scale at Speed is in its second edition and has a 4.7 rating on Amazon, which I’m pretty proud of. More than that, it’s helped thousands of founders empower their teams and release the leadership bottleneck most service businesses fall victim to.
In February this year, Claude Code became really useful and I started to play with it seriously. The first thing I did was build an AI agent-builder, which took me around 16 hours. I launched it with a website built by the agent (moltbookagents.net) and experimented with hooking the agent up to Google Ads and Google Analytics to see if I could successfully get it off the ground. In the second week of the experiment it made around two grand in software sales. I totted up what I’d spent on tokens and ads, and unsurprisingly I’d spent around two grand.
Then GetViktor came along with an engineering team and tons of money and blew my agent builder out of the water. (You can try Viktor with some free tokens here). I retired my agent and switched. I now use Viktor for almost all operational and administrative tasks in my consulting business.
I moved on. And I built this thing that’s been in my head for years – a coffee recommendation app. I’d been lying on a beach in Santorini in around 2014 and had this idea for something that would let me say "I like coffee X and Y", and it would recommend other coffees I might like that were available locally.
Silly story: I’ve been an adjunct professor and visiting lecturer at Hult Business School for years, and I’ve been using the coffee app to illustrate the difference between an entrepreneur and everyone else: Anyone can have a great idea. Only an entrepreneur actually executes it. I said to everyone: if you build it, you will be a real entrepreneur. Anyway, I’ve described this coffee app to hundreds of wannabe entrepreneurs over the years and nobody had done it.
So I sat down with Claude Opus 4.5 and finally built coffeepalazzo.com in six hours one Saturday when it was raining. It’s got 3,000 coffees in it, from the UK, US and Canada. There’s even a Telegram bot.
I felt like I was now up to speed understanding what was possible and my own (not inconsiderable) limitations.
Then Opus 4.6 came out, so I decided to see if I could built an AI version of Scale at Speed. I wanted to give founders, CEOs, COOs and leadership teams a ready reference for solving the most common issues business leaders (predictably) face. Feast and famine. How to scale. The founder bottleneck. Escaping the weeds. And so on.
As a (ahem) leading authority on scaling, there's a wealth of material to draw on. The book. A 160,000-word Scale at Speed course (written and presented by Inna Goncharenko and produced by Lucius Sayles). 3.5 terabytes of workshop videos, and more. All of which I distilled into Scale at Speed AI, powered by Anthropic and built around total privacy for the user.
I launched it to a few friends and some of the 2Y3X team (2Y3X scales agencies programmatically and is what the book Scale at Speed describes).
The response I got was horrible.
So I simplified it. Then I added a voice interface as an alternative to just being a chat interface. I started to experiment with different price points (the voice interface costs real money, whereas chat is low cost and I’m happy to pay for that). Last Monday I had Opus 4.8 optimise the website.
Still not right.
Then Claude’s Fable came out. And it’s infinitely better. You can have a private chat about the problems you’re secretly facing in your business. And you will get industry-leading solutions based on my 32 years of founding, leading and scaling agencies and all the leadership knowledge we’ve accumulated and applied over the last dozen years.
You can even do it as an interactive voice conversation.
Now that it’s finally out in the wild, I am hopeful I’ll get lots of insights about how to improve it – better UI, better UX, more utility.
And this is where you come in: please have a play with it (for free), and let me know if solves your biggest problem in your first session.