The gap
in the loop.
Faster hands, same questions
A Power BI project saved as text files can be edited by a . It reads the files, writes the change, runs the checks and saves a version, often in minutes. Work that took an afternoon of clicking now takes one message.
So building is no longer the slow part. Three questions are left, and the agent can't answer any of them on its own: what should this report do, how exactly should it change, and is the result right?
Seven steps, three players
Every change goes round the same . You set the goal. You write the , with a chat assistant helping you word it. The agent proposes a plan, builds it, runs the and, once you say go, makes a . In between, you open the report and look.
Predict first: in this loop, which link breaks most often?
| Problem | Status |
|---|
Check: skip steps one at a time. Which of these has nothing after it to catch what it misses?
The goal works the same way. A clear spec for the wrong problem passes every check, because every check tests what you asked for, not what you should have asked for.
Found the hard way
The project behind these lessons ran this loop for weeks. At one point every automated check passed, more than seven hundred of them, and the agent reported the build as clean.
Predict first: who found the next problem?
The fix was small: group each tile with its information button, then add a check that every button sits in a group with exactly one tile. That's the loop working as it should. A person finds the gap, and the next check closes it for good.
The rule, in one breath
The agent closes the build gap. The gap that's left is you: the goal you set, the spec you write, the look you take.
None of that is new. It's workshops, specs and testing, the work business analysts already do. This series maps it onto how you a build loop, as the .
Next: you already do half of this.