AI prepares
- Reading and extracting
- Comparing and formatting
- First drafts
People decide
- Angle and priority
- Interpretation
- The exceptions
People own
- Client and stakeholder trust
- Key relationships
- Commitments and final sign-off
Four hours, built to move from how the model behaves to a controlled experiment your team can run on Monday. Interrupt and challenge at any point.
Where AI prepares, where people decide, and where people stay accountable.
Why fluent output isn’t evidence, and where the model is strong versus fragile.
Context, memory, skills and tools — what each one is and when it fails.
The autonomy ladder: choosing the right level of control for each task.
Six management decisions that turn a vague request into a reviewable assignment.
Matching the depth of your check to the stakes — and why agreement isn’t a check.
Worked examples: a messy thread, a launch, and a recurring report, done with AI.
Four tests before you pilot, and one narrow experiment to run on Monday.
A scroll through the core cards we work from — the ideas your team keeps long after the session.
generating · transforming · retrieving · proposing an action
model knowledge · a supplied brief · an approved source · a live trusted source
read · draft · prepare · change · send
factual checking · judgement · exceptions · final action
“The report was issued on 9 May.”
traceable to the brief“The report reached 12,000 readers.”
no source — inventedTell us about your team and we’ll get you set up.