Insights
What we learned building
Notes on e-commerce, custom software and artificial intelligence, written from the projects we deliver.
The business isn't implementing AI: it's solving bottlenecks
Companies without ordered processes fail at implementing AI. What you're actually hiring isn't a model, it's a team that has already solved real market problems.
Selling wholesale and retail in the same store without breaking anything
The same product, two prices, two business logics. How we built a store that serves both audiences without duplicating the catalog or driving the team crazy.
When it's worth automating a process with AI (and when it isn't)
Not every repetitive process needs artificial intelligence. The criteria we use to decide, before writing a single line of code.
Digitizing an industry that never asked for it
Not every sector was already pushing toward digital. How we open an online channel in an industry where neither the customers nor the competition were asking for one.
Integrating payment gateways in Argentina: what nobody tells you upfront
Choosing the gateway is the easy part. The decisions you can't undo later are the ones worth making on day one.
Custom software, or a platform that already exists?
The question isn't which one is better, it's where your competitive edge actually is. A criterion to decide without falling in love with the answer.
Your slow site is costing you sales (and you can measure how much)
Speed isn't a technical metric: it's money. What to measure, what to ignore, and where to start without rebuilding everything.
What to ask your software vendor (even if you're not technical)
You don't need to know how to code to tell if what's being built for you is healthy. The questions that separate a good vendor from one that will leave you stuck.
How to build a chatbot that doesn't annoy people
Most chatbots make service worse instead of better. The design decisions that separate a useful one from one people dodge.
An MVP is not a half-built product
\"Let's ship an MVP\" became the excuse for delivering something incomplete. What it actually is, and how to define one that's worth doing.
