The short answer
Software development runs through six phases: analysis, design, build, testing, release and maintenance. Together they are called the software development life cycle, or SDLC.
Each phase answers its own question. Skip one and the question comes back later, at a higher price.
The six phases
Analysis: what are we solving?
You decide who the software is for, which problem it solves and when it is done. The result is a short description of what the software has to do, including the hard requirements, such as privacy or connections to existing systems.
Design: how does it fit together?
This is where the choices you live with for years are made: how the system is structured, where the data lives, which technology you use and what the screens look like. A mistake here is hard to undo.
Build: writing the code
The developers build the software, usually in small pieces that can each be checked on their own. It is the most visible phase, and these days the fastest one.
Testing: does it do what it should?
Automated tests check that the code works and that existing features did not break. Then people check that the software does what the analysis said, and try to break it.
Release: safely in front of real users
The software goes to the server. Passwords stay out of the code, logins and permissions are right, and backups and monitoring are running. Whatever is missing here, you only notice when something goes wrong.
Maintenance: does it keep working?
Software is never finished. Security updates come in, users ask for new features and the systems around it change. Over its whole life, maintenance usually costs more than the first build.
Waterfall or agile: the same phases, a different rhythm
The phases are always there. The way of working decides how often you go through them.
With waterfall you go through them once, in order: all of the analysis, then all of the design, then all of the build. That suits a project where you know up front exactly what it has to become. In practice you rarely do.
With agile you go through every phase in short cycles of a few weeks. You build a small piece, test it, release it and learn from how people use it. The next cycle’s analysis starts with what the last one taught you. Most software teams work this way, and so do we.
What AI changes in each phase
Now that AI writes much of the code, the build is no longer the longest phase. The work shifts to the phases around it.
Analysis matters more
An AI model fills every gap in your brief on its own, often not quite the way you meant. A good description up front is now the most important piece of work in the whole project.
Design stays human work
A model picks what works today. An architect picks what still works in three years. Choices about structure and data belong with a person who can account for them.
The build gets much faster
In our own work, building with AI delivers up to ten times as much as typing every line ourselves. What used to take weeks now takes days.
Testing becomes the brake you need
The faster code arrives, the more you depend on the checks that stop mistakes. A model almost always thinks its own code is fine.
Release is where it goes wrong
Security, permissions and backups are what AI skips most easily. Plenty of quickly built apps work fine until the first real users show up.
What that looks like step by step, from the first description to reviewing every change, is in how to develop software with AI.
AI made the build cheap. The phases before and after it now decide whether your software is worth anything.
Where projects get stuck in practice
Most problems we run into are not in the build. They are in a phase that got skipped.
An app built with AI in a few weeks works in the demo but never had a proper release: no backups, passwords in the code, everyone can reach all the data. That is the last 20 percent that turns out to be most of the work, covered in the last 20 percent of a vibe-coded project and getting a vibe-coded app production ready.
A system that has run for years was once built without a design. Every new feature takes longer than the last, because everything is tied to everything else. Another developer will not fix that. Someone who redraws the structure will, which is what our strategic tech consulting is for.
What are the phases of software development?
What is the SDLC?
What is the difference between waterfall and agile?
Which phase of software development takes the longest?
Can you skip phases when you build with AI?
Want software built without skipping phases?
We build with AI and keep the analysis, the design and the review in our own hands. You get the speed of AI and software you can still maintain a year from now.
