What it means in practice
A developer stops being the person who writes the lines. They become the person who designs the system: what the agent needs to know, what guardrails it has, how you verify it did good work, and when a human has to decide.
The goal we set in the workshops is deliberately sharp: developers write not one line of code manually. Not because they cannot, but because their time is worth more designing the system than typing.
What it is made of
The road from prompting to engineering runs through four questions:
- How do I ask the model? The basic work with a prompt.
- What does the model need to know? Context engineering.
- How exactly do we describe what should be built? Spec-driven development.
- How do we design an environment where agents work reliably? Harness engineering.
Why management needs to know this
Two things change at once. First, the man-day stops making sense as a billing unit. In 2024, writing code by hand was around 70 percent of the work and hours were what got billed. When agents run in the background, you pay for specifications, agent time on a server and tokens, not for hours in a seat.
Second, the spread in team performance widens. One developer who has taken agentic engineering on board can out-produce the rest of the team combined.
How we teach it
The full-day Agentic engineering workshop is run by Michal together with Josef Nevoral, a CTO who builds software this way daily. The work happens in your code, not in a practice project.