If it itches, scratch
There's a very weird thing that happens when you do something without a goal. Also, it's fun.
I've had more time and mental capacity the last few months, and alongside riding my bikes more, I also explored how AI can help me work differently (maybe even better?!).
We had already been using AI (both machine learning and LLMs) heavily at the company to extract and enrich data, as well as to write code, but besides that our experiments on the operations front hadn't gone much further than recording and summarizing meetings, and helping to write an email here and there.
Something bothered me about the fact that I was using this world-changing technology to summarize long text. It felt ambitionless; and kind of boring.
So here's some examples of stuff I've worked on in the last few months that made my work more fun (and as a bonus, made me more productive too).
I've always loved building new products. It's creative, has a huge margin for error, and the potential is uncapped (mostly because you know almost nothing yet). At the same time, I've kind of icked at the idea that I now do "sales".
I love the discovery of a client's problem deeply and (only when I can actually help) actually sitting down and designing a solution, battle testing it, and then helping them present it internally. People light up when you offer value to them before they commit to anything, and honestly it just feels good to do it.
Naturally, I tried to come up with a way to minimize the part of sales I dislike (qualifying leads, checking in, follow-ups to reengage customers that got busy) and maximize the time I spend building stuff with (and for) customers.
My solution was a claude-code based sales agent which reads my emails for a list of customers weekly, does deep research on what their team or company has been working on and prepares a summary and drafts for me to follow up. Drafts often need to be modified to sound less robotic, but the rest works incredibly well and has helped me be more informed and present with each customer.
If you're a nerd like me, you can read more about the architecture behind this tool.
What even is product management? In my experience, if you're building a new product, you don't need a PM, especially in a startup. The vision is aligned across the team, market feedback flows through 1-2 people and prioritization just falls into place. You anyway usually have less resources than you'd need, so planning and optimizing what the team spends time on is rarely a topic.
As products grow however, I've found myself having to balance different people pushing for what they think should be prioritized, as well as a range of opinions about "what the market needs". Essentially mixed signals, and a blurry view on where the opportunities lie or what our own internal vision regarding our product is within the market. We are of course not the first nor the last team to be feeling this.
Now that we can build anything in a fraction of the time (at least in software), the real question becomes what should we build?
To deal with that, I've found that developing a framework of what matters for your product really helps in then letting AI agents process and map the flood of information to each section. That leads to a (surprise surprise) live dashboard which — weirdly enough — feels so custom and on point that I actually want to look at it.
I thought this might be a nice product idea (probably because right around when I got a first working version I saw YC's request for startups essentially pitching this to new founders) and built a prototype. It turned out to be too expensive at the time, so I let it go (you can see the framework in an overly fancy visualization).
My very own "cursor for PMs" though is very much up and running, and the fancy dashboard is nothing other than a self-contained HTML file.
Things deserve to be beautiful. Yes, we live in a constrained world and yes valuable things don't always shine. But shiny things always shine.
I grew up in a family that cares about aesthetics and design. And in some ways it's a curse, because my supposed "taste" makes me really uncomfortable when I see things that could look better. A beautiful apartment badly furnished. A high-ceiling gym stuffed with ugly and unnecessary equipment. A slide deck for a customer with all the right messages that hurts the eyes.
Luckily, the last example can be solved with a few .md files and some attention to detail. I've been obsessively refining slides and skills files. A couple of examples I'm proud of are my Agent Economy Zurich slides and the Studio template.
If you're interested in more templates you can use, have a look at this collection.