About
I built this because
I needed it.
Customer Magic is made by one person who has spent fifteen years doing this job for other companies, and now does it for his own — badly, on the days it competes with everything else on fire.
I'm Ben. For about fifteen years I was the growth co-founder — across roughly ten startups, running marketing, sales and product-led growth as one craft rather than three departments. That is the whole of my working life: figuring out where a company's next revenue actually comes from, and then getting someone to actually go and get it.
In early 2025 I taught myself to build software properly, and shipped a paying SaaS in about three months. Which meant that for the first time the two halves were in the same pair of hands — the growth judgment and the ability to encode it.
Since then I've been doing it in public. I co-founded Pirate Skills in Cologne, where we teach small teams to build and grow their own software with agents as an actual team rather than a faster set of tools — and where we ship a real product every quarter, because teaching from memory goes stale within about a month. I run a monthly in-person meetup here, speak at things, and publish every week.
Then I ran into the problem this product exists for, in my own business. I connected everything to my agent in an afternoon — inbox, calendar, CRM, Stripe — asked it who I should contact this week, got a genuinely good answer, and contacted nobody.
Not a discipline problem. A missing piece of furniture. The agent could find things and had nowhere to put them, no opinion about what came first, and no memory of what I'd already decided. So every morning I re-derived the same list and did nothing with it again.
Customer Magic is the workbench that was missing. It sweeps every source you have already connected, sorts what it finds into six plays, ranks them with the reason attached, and drafts the first move. I built it for myself first, and I run my own businesses on it.
Automate the labour. Protect the understanding.
Every competitor picks a side. Pure automation creates dependence and plateaus; pure teaching creates homework nobody does. We do both at the moment of the action — the work gets done, and the reasoning behind it stays visible so you keep it.
The problem is order, not discovery.
Founders don't fail at sales because they can't find the opportunities. They fail because twenty possible moves fan out with no order between them, the hour ends, and nothing was sent. Solve the order and you solve the starting.
You should end up sharper, not more dependent.
If you cancel after a year you should be a better operator than when you arrived. A tool that makes you worse at your job the longer you use it is a tool with an expiry date.
It's for you if
- You already work inside an AI agent every day and have connected at least one real source to it.
- You have a real book already — live customers, accounts that could be bigger, a few hundred names who bought once.
- You personally own the number, whether that's as a founder of a 1–5 person team or as the person who owns pipeline for five or six reps.
- You want the hour to end with something sent, not with a tidier CRM.
It isn't, yet, if
- You're brand new with no relationship history. There's nothing to sweep, and we'd be selling you a search over an empty room.
- You don't already work in an agent. The whole product assumes that, and pretending otherwise would waste your money.
- You want a CRM. This is not one — it reads yours, and gives it back to you ranked.
Saying this out loud costs us some signups and saves both of us a bad month.
Customer Magic operates under Pirate Skills GmbH, Cologne, Germany — the company I also run Pirate Skills through, founded in 2016. They're separate products with separate customers; the shared entity is a corporate arrangement, not a shared mailing list.
Full legal details, VAT identifiers and the register entry are on the imprint. Where your data goes is on the security page.
Easiest way to judge whether I know what I'm talking about
Read one, then decide. Everything I claim about how this works is argued in public every week, where it can be wrong in front of people.