Widen Our Aperture

This is the mode most people mean by “AI for research”: do more, faster. Compile desk research, translate sources, ingest messy notes, expand what a constrained plan would normally allow.

A bigger workbench — then an intentional narrow.

The value is a bigger workbench — not endless exploration. Widening without a later narrowing is how teams drown in fluent summaries and call it insight.

More data and deeper understanding are not the same thing. Constraint is still part of the craft.

Core principles

  1. Use when the field is still too small, sources are scattered, or volume would exclude useful context.
  2. Do not stay here to avoid talking to a person, or when output leaves your desk without verification.
  3. Every claim needs a source label. If Copilot cannot label it, it does not belong in the brief.
  4. Never invent stakeholder opinions, quotes, or market numbers.
  5. Exit with Keep / Park / Drop — if you cannot name what you are dropping, you have not finished.

Applied recipe: Client Research & Meeting Prep

Check yourself

What finishes a Widen aperture session?

Do this in Copilot

Run structured desk intake on your live task (or a safe sample). Produce Keep / Park / Drop. Bring the narrowed set into Lesson 3.

Send this first. Paste approved sources in the next message. Follow data-privacy rules.

Structured desk intake

I am doing desk research for [PROJECT / DECISION]. When I paste sources in my next message:
1. Extract claims, facts, and open questions. Label each with its source.
2. Flag gaps, contradictions, and anything confident but uncited.
3. Do not invent stakeholder opinions, quotes, or market numbers.
4. Ask me up to 5 questions that would most change what I should gather next.
5. Then propose a narrowed focus: what to deepen, what to drop, and why.
Do not draft a final brief yet. Reply "Ready — waiting for sources."

Techniques: Source-labeled intake · Constraints

  • After intake, force Keep (3 must-stay items), Park (dated, with why), Drop (noise/duplicates).
Open Copilot

Did you run this in Copilot? Mark complete when you have tried it.

Next lesson: Help Us Notice Nuance →

AI

Part of the MillerKnoll collective

Studio · 2026