What genuinely goes
AI absorbs one very specific kind of work well: the kind whose result can be checked quickly. A first draft, converting one format into another, searching a large pile of documents, repetitive code you can eyeball for correctness in seconds.
Notice what those have in common. They are not the hard tasks — they are the long and uninteresting ones. The work you were paid for by the hour rather than for your judgement.
In software testing that means: writing the third variation of a test case, filling in a tracking spreadsheet, reformatting a defect report. Nobody chose this profession for those three things.
What does not go
Three things hold, and none of them is technical:
- Deciding what is worth doing at all. A model answers the question you asked. It does not tell you it was the wrong question. In testing, knowing what you will not verify — and owning that call — stays entirely human.
- Recognising an answer that is plausible and wrong. This is the real risk and it is under-discussed: AI is confidently incorrect. You have to know the subject to catch it. A beginner who delegates without being able to check does not save time — they cost it.
- Carrying the responsibility. When a release ships, somebody answers for that decision. No tool takes that seat, and European AI regulation now writes it into law: certain decisions must remain under human oversight.
What we are not going to tell you
That nothing changes would be false, and you would know it.
What changes is how your time is distributed. Someone who spent 60 % of the week on repetitive production and 40 % on analysis will watch those proportions invert. For most people that is good news: at the same hours, productivity rises because the time moves to work that carries value. For anyone whose role consisted only of the repetitive half, it is less good news — and saying so is more useful than hiding it.
Those are exactly the people we are most useful to. Not because we sell insurance against the future, but because the missing skill is learnable, and it takes months rather than years.
What separates the people who gain
Across our cohorts the difference is not the starting technical level. It comes down to two habits:
Asking precisely. A vague request produces a vague answer, which then takes longer to fix than doing the work yourself would have. This is where most people conclude "it doesn't work" and stop.
Checking properly. This is the skill gaining the most value, and it happens to be the tester's craft — judging whether a result is correct, and against which criteria. That profession was always methodical doubt; AI did not change its nature, it raised the demand for it.
What ADC actually does
We do not teach "AI" as a separate subject. We teach professions with AI already inside them, because that is how it shows up at work: not as a tool you open, but as a different way of doing the same tasks.
- Our tracks — from manual testing through automation, with no technical prerequisite to start.
- ISTQB® CTFL — the qualification recruiters filter on, included rather than sold separately.
- Consulting, if the question is about your team rather than your own career.
The same question, from the other side of the desk
Leaders ask us the mirror version: "does AI let me run a smaller team?" What we observe is almost always no — but it does let the same team hold a wider scope, which is not the same thing and is usually worth more.
And what blocks that is never the AI. It is the absence of clear criteria about what must be verified at all. A quality audit establishes them; a test strategy writes them down.