On this page
The problem
Most small businesses don’t run on a system: they run on Excel spreadsheets, WhatsApp threads, emails and manual checks. It works while the business is small. Then it starts to hurt: the information lives in five places, every process depends on one specific person remembering to do it, and a mistake goes unnoticed until it has already cost money. The day that person goes on vacation, the business finds out how much it depended on them.
Faced with that, the way out usually offered is a SaaS: you rent a generic tool that almost fits your operation, and you pay for that “almost” every day by adapting your business to the software instead of the other way around. The third option, a focused system of your own, gets ruled out for fear of over-engineering. In my opinion that’s a mistake: a small, well-built system organizes your real operation without depending on third parties, keeps your data centralized in one place, and it’s yours.
What I do
- A system that fits your operation, instead of the other way around. First I understand how your business works from the inside: what’s done by hand, which spreadsheet each piece of data lives in, where the bottleneck is. The software adapts to that.
- I organize the manual work: processes with clear steps, checks that used to depend on someone remembering, reports that build themselves instead of being put together at the end of the month with copy and paste.
- AI where it adds value, not because it’s trendy: repetitive, structured tasks (classifying, extracting data from documents, routing, summarizing) are the ones that give the best return. The rest is handled by good software, without hype.
- Yours, with no lock-in: the code stays in a repository of yours, documented, so that another developer can take it over. I don’t tie you down; you stay if it suits you, and you can leave whenever you want.
- Built with standards that can be audited: a typed backend (FastAPI, Python), an orderly database (Postgres), tests on the critical logic and deployment that rolls itself back if something fails. The stack lives in the code I hand over and it can be audited; I name it for that reason, not to show off.
Integrations that talk to what you already use
A system of your own is of little use if it forces you to keep copying data by hand from somewhere else. For simple connections a workflow tool like n8n is enough, and I wrote separately about when it falls short (in Spanish). What follows is what gets written inside the system. These are the integrations I build, and what each one solves:
- API integration with the services you already have. If the provider publishes an API, it gets connected: payment gateways, couriers, ERP, CRM, spreadsheets, email, WhatsApp. With retries, a log of every call and an alert when the other side goes down, which is what separates an integration from a script.
- Electronic invoicing with the SII (Chile’s tax authority). Issuing and receiving electronic tax documents from your system: the data comes from your operation, instead of someone typing it again into another platform. It includes managing the folio numbers the SII assigns, the status of each document and the record of what was sent and what was rejected, so you can audit what happened with each one.
- Bank reconciliation. Matching transfers against what your system expects to collect, with doubtful cases set aside for human review instead of being forced through. It’s what runs at CM Vida Plena.
- Inventory and data across systems. Stock, prices or catalog synced between your operation and the store, like the per-branch stock at PetShop MG.
- Migration of what you already have. Your spreadsheets and your historical data go into the new system; you don’t start from zero or lose the history.
- Single sign-on and role-based permissions. Each person sees what they should, and there is a record of who did what.
If your provider has no API and only delivers files, that works too: the file gets processed. What I don’t do is promise an integration without first seeing the documentation on the other side.
What’s out of scope: if what you need is a website or a store, that’s WordPress development, not a system. And if you already have a system and only want to add AI to it or automate part of it, that’s AI process automation. I tell you which case is yours before quoting anything.
Real case
Flagship case: Console CM VidaPlena. A medical center that ran its management on spreadsheets and manual checks now runs on its own system built with FastAPI and Postgres: its processes organized in one place, with role-based access control and daily encrypted backups. It works in production, they use it every day, and the operation stopped depending on third-party services: its data lives centralized, in a system it owns.
And the same idea on a smaller scale, inside a website: the Comercial Moya quote tool processes quote requests every week, real business logic running in production since its launch.
How I do it
- 01
Discovery
I understand how your operation works from the inside: what's done by hand and where the bottleneck is.
- 02
Build
I build the focused system that fits your operation, with standards that can be audited.
- 03
Handover
The system in production, documented, with the code in your repository.
What you get
- The system running in production, past the prototype and demo stage.
- The code in a repository of yours, documented and transferable. No lock-in.
- Your operation, which used to live in spreadsheets, organized in one place.
- An orderly handover: I explain how it works and how to maintain it, so you don't depend on what is in my head.
What exactly is custom software?
It's a system built for how your business works, instead of forcing you to adapt your way of working to generic software. Instead of renting a tool that does 'almost' what you need, you build the one that does exactly what you need: your processes, your rules, your data. And unlike a SaaS, the result is yours.
How much does custom software cost?
It depends on the scope, so I don't work with a list price. The value lies elsewhere: you stop depending on third parties for your daily operation, the system is yours and your data is centralized in one place instead of spread across other people's tools. That's why the first step is a free diagnosis, where I review what you use today and tell you whether it makes sense to build, or whether organizing what you already have is enough.
Custom software or off-the-shelf software?
It depends on the problem, and I tell you straight. If what you need is standard (accounting, invoicing, a common CRM), off-the-shelf software comes out cheaper and faster. Custom wins when your operation has something of its own that no generic tool covers well, or when you already pay for three or four subscriptions that only half fit. The question is which one fits your case, more than which one is better in the abstract.
How do I know if I need a custom system?
Typical signs: your operation lives in spreadsheets that break as you grow, you depend on one specific person remembering to do things, or you pay for several SaaS tools and none of them quite fits. If none of that happens to you, you probably don't need one, and in the diagnosis I tell you so without selling you a development project.
What happens if tomorrow we stop working together? Am I locked in?
No. The code stays in a repository of yours, documented, and any competent developer can pick it up from there. I don't tie you to me, which is exactly the opposite of the SaaS you rent and can't take with you. The honest part is that a custom system needs someone technical to maintain it, since you can't manage it on your own the way you manage WordPress; that's why I leave it documented, so that person doesn't depend on what is in my head.
Does this include artificial intelligence?
Where it really adds value, yes. For example, AI can read a document and extract the data instead of someone typing it in by hand. But I don't add it because it's trendy: if good regular software solves the task, it goes without AI. And if what you're looking for is adding AI to a system you already have, that's a different job: AI process automation.