Behind the course
The story behind this course
By Mercedes Perez-Capilla · May 2026
Product Lead, Financial Services
I built this course in December and January, before any official materials existed, using Claude skills I wrote myself.
The first is the automation-opportunity-finder. It's the skill you build in Module 1 to find the right automation to start with in your own work. The other two built the course itself. The course-formatter made every lesson come out structurally identical, the same shape and rhythm with no drift, and the universal-example-layer kept one voice and one person, Riley, consistent all the way through. Early drafts had a different example in each lesson and it didn't hold together. One decision fixed it: one person, one story, all the way through. The skill kept me honest.
All three are open-source on my GitHub if you want to use them.
So when I say this works for real knowledge work, I'm not working from theory. The course is its own case study.
I also built a command, /mercedes-lab-update, that re-checks every lesson against Anthropic's current Claude Code documentation and flags anything that has drifted out of date. This course doesn't just teach automation. It's kept current by it.
Why Riley
I needed a protagonist because professionals don't learn from abstractions. They learn from someone like them hitting the same wall they're about to hit.
Riley Harper is an Operations Director at a mid-sized financial services firm. Fifteen years of experience, no coding background. Her Monday morning routine: pull margin data from three systems, format it into a report, add commentary, email it to four people. Four hours, every single week. By the end of Module 1, it takes 18 minutes.
I made her specific on purpose. IM/VM, Net Exposure, Margin Call Status. Real numbers, like the £13,875 a year she saved from one automation she built herself. I wanted “Riley's margin report” to stick in a way that “your repetitive task” wouldn't. Every quote from her is something I've actually said, or heard said in an ops meeting.
On the terminal
The course promises you can automate your work without writing code. That's true. And Claude Code involves a terminal.
I'm not going to pretend the terminal doesn't exist, because the moment it stops being scary is the moment the whole thing opens up.
Here's the honest version. I came from Visual Studio, a proper editor with menus and buttons and things to click. When I first saw a bare terminal, just a blinking cursor, it felt like a different world. So if it looks that way to you, I understand completely. But the terminal isn't a coding environment. You describe what you want in plain English and Claude builds it. You're not typing syntax, you're giving instructions. The moment one sentence turns into a working automation, the fear goes away. There's a desktop app now too, so for parts of this course you won't need the terminal at all.
Who this is for
I've deployed systems to production in regulated financial services environments. Riley's world is collateral and margin operations because that's the world I know and it makes every example specific. But the same principles apply in legal, marketing, consulting, operations, healthcare, education. Repetitive knowledge work looks the same across industries. So does the relief when it stops being manual.
The lesson that matters most
Lesson 4.5 is called “When to Keep a Human in the Loop” and I think it's the most important lesson in the course.
Most AI content treats human oversight as an afterthought. It isn't. It's a design decision you make upfront. Automation does the work. Autonomy makes the decision. Those are different things, and knowing which is which is your job.
The test I use is simple. If this runs while I'm not watching and produces something wrong, what's the worst that happens? If the answer involves a client, a regulator, or anything irreversible, the human stays in the loop. That's not a limitation. That's judgment.
An honest note on timing
I built this in January 2026, before there was much to learn from. There was a handful of videos on how to build a single skill, and the fastest way I know to understand something is to teach it. The landscape changed while I built. The Anthropic courses are now free, official, and excellent, going from the foundations all the way to the more technical side, and they should be your first stop for the tool itself. I've done them myself and they're worth your time. This course points you there.
What it still does is show how a non-technical business leader actually applies Claude Code to her own work, with a real ROI and a governance instinct attached. I built into a gap, and that gap is now closed. Saying so plainly matters more to me than pretending otherwise.
Where this goes
The outcome of this course is simple. You go from watching AI change everything to having built an automation that saves you real hours. Done is the credential.
What matters most is that you start today. Build something. Try things out. Learning by doing is how this really works. And let me know what you build, whether it's an automation that makes your work or your life better, or a creative idea you bring to life. You can build anything. The sky is the limit.
What it was worth
To me, more than the course itself: I came out able to build and ship with Claude Code, not just talk about it.
“I learnt markdown and built my first skill to run my daily standups. Now I'm using Claude Code and doing the CCA. As a starting point, it was the rocket launch.”— Eric, Technical Lead
Work with me
Want to build something specific?
For teams navigating AI adoption, or individuals who want to move faster with a guide, book 30 minutes and we'll figure out what to build next.
Book a session →