Scratchpad

Contextual Assistance — Smart Book Builder

Testing how to introduce powerful AI tooling at the moment it becomes relevant — without blocking momentum or hiding real complexity.

The problem this explores

Admins often assemble board books manually:

  • uploading documents one by one
  • creating tab structures
  • linking agenda items by hand
  • checking for missing sections

In many cases, they’re already paying for tools that can help — like Smart Book Builder — but those tools are hidden behind buttons, unfamiliar workflows, or assumptions about readiness.

This experiment explores a different approach:

What if the system noticed how someone was working, and introduced Smart Book Builder only when it was clearly relevant?


An important constraint

Smart Book Builder isn’t “one-click magic.”

It involves:

  • templates
  • configuration
  • previews
  • human judgment

So the goal here is not to automate everything, but to:

  • reduce setup friction
  • pre-fill what the system already knows
  • hand off into the full tool at the right moment

Two scenarios explored

This prototype supports both:

  1. Smart Book Builder enabled

    • The system prepares a draft starting point
    • The user reviews before applying anything
    • Complexity is acknowledged, not hidden
  2. Smart Book Builder not enabled

    • The system acts as a promotion
    • It shows what the user is doing manually
    • It quantifies time saved instead of overselling AI

In both cases, the emphasis is on respecting momentum.


What to look for

  • Does the right-side panel feel like a bridge or a speedbump?
  • Does showing “prefill evidence” earn the system the right to intervene?
  • Does the promotional version feel informative rather than salesy?

This experiment intentionally separates:

  • intent (clicking Smart Builder)
  • preparation (contextual panel)
  • execution (the actual builder)

Explore the prototype

Open the Context Assist — Smart Book Builder experiment
/now/ca-bookbuilder

This is an exploration of timing, not feature completeness.

If the panel feels unnecessary, that’s a valuable outcome — not a failure.