Main idea

Research carried out in successive iterations expands on its own: each source opens questions, each question brings in concepts, each concept justifies more reading. Without a boundary, the loop never ends.

So it starts with an explicit mission: a central question, what is in scope, what is excluded, the themes to understand, the kind of knowledge expected. To prepare for Adobe Summit Paris 2026, the mission was limited to five themes — content production in the agentic era, AI-driven brand governance, queryable document systems, visibility in LLMs, conversational interfaces.

Convergence then only exists relative to that scope. You can never conclude "we know everything about AI agents"; you can conclude that, for this mission and within this scope, a new iteration is now unlikely to substantially change the model.

Why it matters

The scope isn't red tape added to the research: it is what makes a stopping criterion possible to state. Without it, "searched enough" has nothing to refer to.

It also changes what you write at the top of a research brief: the exclusions matter as much as the subject.

Nuances and limits

A badly drawn scope makes the research converge on a model that is coherent but beside the point. Convergence attests to stability within the frame, never to the soundness of the frame.

And a scope that is too narrow converges fast for the wrong reasons: it rules out in advance whatever could have contradicted it.

Open questions

• On what signal should research in progress widen its scope rather than converge within the one it was given?