
Vibe coding at companies: the prototype takes an afternoon, the hard part comes after
When anyone can build an app, the question changes.
Services
AI Solutions DevelopmentCustom AI TrainingsAI Workspace SetupIndustries
ManufacturingEcommerceHR AgenciesForeign TradeAccounting FirmsMarketing AgenciesInsuranceHealth ServicesLegal & NotaryReal EstateFinancial ServicesLogisticsThe price of building dropped a lot. What comes after, much less so.

For years, the custom software question was almost only about budget: building your own was expensive and took months, so most companies adapted to off-the-shelf systems and filled the gaps with spreadsheets. With AI, building costs much less. The useful question now is a different one: which part of the business should be your own, and who will keep it running afterwards.
It's a system developed for one organization's processes, instead of a product designed for thousands of customers. It can be big, like a sales platform, or small: a portal where suppliers upload their documents, a dashboard that brings together data from three systems or an internal tool that replaces a shared spreadsheet nobody understands anymore.
Off-the-shelf software is bought or rented ready to use. It handles well what is the same everywhere and forces adaptation where things differ.
AI tools have become part of the daily work of people who write code. According to Stack Overflow's annual survey, half of professional developers already use them every day. In practice, a first version that used to take months can now be running in weeks.
The same survey shows the other side: the most common complaint is dealing with AI solutions that are "almost right, but not quite." Writing code got faster; reviewing, deciding and maintaining what gets built still takes people with judgment.
For a company, the important change is that many projects that didn't add up because of cost now do. That opens an opportunity and also a risk: when building is easy, it's easy to build too much.
A simple rule: buy what is the same everywhere and build what defines how the company works.
Accounting, payroll, email or e-signatures rarely justify custom development. Mature products exist, and the advantage of building them by hand is minimal.
Custom software usually makes sense in three situations:
When the process is how you compete. The way you quote, serve a certain type of customer or coordinate a particular operation. Forcing it into an off-the-shelf product ends with the system dictating how people work.
When the work lives between systems. If the team spends hours copying data from one screen to another or keeps parallel spreadsheets to cover what the system doesn't do, there's a custom piece waiting there.
When you want to test something new. A service, a channel or a business unit that doesn't exist yet. A small version of your own lets you learn fast without committing to an annual license.
It's rarely all or nothing. The most common path is to keep the core system and build around it: connect what already exists and develop only the part that makes the company different.
Cheaper development doesn't make what comes after free. There are three costs worth looking at from day one.
Maintaining it. All software changes along with the business. Someone has to be able to understand and modify it two years from now, so it matters that it's documented and that the company owns the code.
Running it. A custom solution has to be cheap to run, easy to change and secure from the first access to real data. That technical challenge is still central; it has just changed shape.
Getting it used. Many projects work well and still end up half used because nobody changed the way of working around them. Training each role and measuring usage in the first weeks matters as much as the code.
In Ambolt's projects, what best organizes these decisions is starting with the process that hurts the business most and putting a first piece into use within a few weeks. Once something is running, the conversation about what else to build becomes concrete.
With AI, custom software stopped being a luxury for large companies. That changes the underlying question: when building is cheap, the valuable part is choosing well which piece of the business to own and who will keep it running. Those who make the most of this moment will probably be the ones who can say no to what is better bought and focus their own development on the few things that truly set them apart.
It's a system developed for one organization's processes, unlike off-the-shelf software, which is a standard product designed for many customers.
Off-the-shelf software is bought or rented ready to use and handles well what many companies have in common. Custom software is built for a specific process and adapts to how the organization works, but it requires maintenance and someone to keep it running.
When the process is part of how the company competes, when work gets lost between systems and spreadsheets, or when a new service needs to be tested with a small version. For standard functions, like accounting or payroll, buying usually makes more sense.
It greatly speeds up writing code, so a first version can be running in weeks. Reviewing what gets built, deciding what to do and maintaining it over time still require business and technical judgment.
Customer or supplier portals, dashboards that integrate data from several systems, internal tools that replace shared spreadsheets and modules that connect to an existing ERP to cover a particular process.
Maintenance and changes as the business evolves, infrastructure to run it, data security and the work of training the team to use it.
About Ambolt. Ambolt is a product agency that helps companies in Latin America decide what to build and put it to use fast, with no minimum contract term.
We specialize in finding areas to optimize your company's operations, making them leaner, faster and more productive.
Talk to us →