Training programme
Building software with AI: learning the method, not the tool
This training is for software teams. It shows, on the team’s own project, how to get AI to write code and how to check whether that code is right. It does not promote a particular tool — tools change, the method stays.
Getting code written is easy. Knowing that the code is right is the hard part.
The questions it answers
The programme is built around three questions
- 01
The team writes code faster with AI, yet the review queue keeps growing — what has to change?
- 02
How do you measure whether generated code is correct, and rule out “it looks like it works”?
- 03
Which work goes to AI and which does not — where is the line drawn?
Programme
How it runs
01
Setting up the context
Output quality tracks the context you provide. If project rules, architectural decisions and the list of “never do this here” are not written somewhere durable, every session starts from zero. In this part the team builds those files for its own project.
02
The development loop
Request, plan, small steps, run it and look, correct. Keeping the loop short instead of asking for everything at once, and leaving something that actually runs at every step. The team practises this on its own codebase.
03
The discipline of verification
This is the backbone of the programme. How every output gets tested: with tests, with measurements, against real data, and with your own eyes. “Probably right” is not a result, and an unverified output costs more than code never written.
04
Team setup and handover
Reviewing AI-assisted work, keeping a record of it, and making sure it can be handed over inside the team: who approved what, where each decision is written down, and where a newcomer starts.
Scope
Who it is for, and who it is not for
For this
- A software team already using AI but unsure how far to trust the output
- A team that wants to work on its own codebase, under its own rules
- Teams whose speed-up disappears under the weight of review
Not for this
- Anyone expecting a screen-by-screen demo of one tool — tools change, this teaches method
- A team that does not write code; this programme is for developers
- An organisation hoping AI will write the software with nobody looking at it
Questions
Questions about this page
Which AI tool does this training teach?
The programme is not a tool demo. It runs with whatever tool the team already uses; what is taught is how to set up context, keep the loop short and verify the output. The method survives a change of tool.
Does the training run on our own codebase?
Yes, and that is the preferred way. A sample project explains the concept but hides the team’s own constraints: real architecture, real technical debt and real review habits only show up in your own code.
How is the correctness of AI-generated code measured?
A checking regime is built per type of output: tests where tests apply, measurements where something can be measured (time, size, query results), and a look at a screenshot for visual output. Nothing unchecked counts as done.
Is the training delivered remotely?
Either way works. What matters is that the team can reach its own development environment: the programme is hands-on, and participants work on their own projects on their own machines.