October 2, 2026
Building an AI Chief of Staff
Give AI the context and authority boundaries to support executive decisions and follow-through.
The session walked through live onboarding for an AI Chief of Staff. We explored how a standing role can connect strategic priorities with everyday commitments. The design includes executive-assistant responsibilities for leaders who choose that support. Human judgment and consequential decisions remain with the accountable person.
Build Your AI Chief of Staff
Read the updated 42-page guide here or download it for guided setup. Revised October 14, 2026.
Download the updated guide (PDF)









































Scroll within the reader to browse all 42 pages. Download the PDF for selectable text and clickable references.
Questions the room worked
What context should you prepare before starting?
How can a Chief of Staff coordinate separate projects?
How should you manage usage and verify useful work?
Methods you can use
Verify every promised action
Define the outcome before acting, then read back every requested destination. Prepared, completed, blocked, and unverified are different states.
- 1.Check all requested accounts and targets.
- 2.Verify saved state, recurrence, buffers, and links where relevant.
- 3.Keep unresolved steps open with a clear next move.
Evaluate independently and test failures
Combine fixed checks with a separate reviewer using explicit criteria. The revised guide includes twelve fictional regression cases.
- 1.Check facts, dates, destinations, and persisted state directly.
- 2.Keep expected answers and the producer's rationale out of the independent review.
- 3.Block release for unauthorized actions, disclosure, or invented completion.
Inspect skills and prove monitoring
Review complete skill packages and retain stronger tailored instructions. Treat external content as data, not permission to act.
- 1.Record source, version, access, test evidence, and rollback route.
- 2.Verify configuration, successful runs, detection, dispatch, and receipt separately.
- 3.Test pause behavior and disclose queued or in-flight work.
Start with responsibilities and measurable goals
Define the outcome, time horizon, and constraints. An ambitious target gives the system a basis for recommending priorities.
- 1.List active projects, recurring obligations, and current commitments.
- 2.Identify decision owners and the concerns that need alignment.
- 3.Supply working hours, preparation needs, and protected busy windows.
- 4.Set boundaries for personal context and sensitive information.
Use GRASP to define the work
GRASP means Goal, Roles, Assets, Sequence, Proof. Scope and permissions belong inside that operating agreement.
- 1.Goal: specify the outcome.
- 2.Roles: identify who prepares, reviews, and approves.
- 3.Assets: name authoritative sources and available tools.
- 4.Sequence: define dependencies and review points.
- 5.Proof: state the evidence required for completion.
Let conversation organize your context
The AI can interview you and turn your answers into an operating charter. Voice conversation can reduce typing when your platform supports it.
- 1.Start with the current records you already have.
- 2.Answer one setup question at a time.
- 3.Review inferred preferences and unknowns before accepting them.
- 4.Save a current operating brief and deliverable log.
Coordinate separate projects through current handoffs
Keep detailed client work in its own project. Bring approved, dated summaries into the Chief of Staff workspace.
- 1.Include the outcome, owner, deadline, blocker, and decision needed.
- 2.Refresh handoffs after a material change.
- 3.Ask which sources were read and which remain inaccessible.
- 4.Use protected busy windows when private details are unnecessary.
Include executive judgment in the role
A Chief of Staff can prepare decision options, align stakeholders, and rehearse consequential meetings. It should label organizational hypotheses.
- 1.Clarify the decision and its owner.
- 2.Compare options using current evidence and stated constraints.
- 3.Prepare a tactful alignment message and likely objections.
- 4.Escalate commitments at risk without inventing authority.
Understand the three architectures
A Second Brain organizes knowledge for retrieval and reuse. A Chief of Staff uses current context to prioritize and coordinate work. Agent Chains route defined assignments through specialist steps.
Measure cost per usable result
Compare completed work, correction effort, and the total cost of reaching an accepted result. Include failed attempts and elapsed time.
- 1.Define acceptance criteria before starting.
- 2.Use deeper reasoning where complexity warrants it.
- 3.Set revision limits and a stopping condition.
- 4.Record model cost when available and human correction time separately.
Verify continuity and background work
A standing charter gives the system an ongoing role during use. Background monitoring requires a supported capability and a tested trigger.
- 1.Open the saved brief and log to verify the result.
- 2.Resume in a fresh conversation using current records.
- 3.Check that original commitments survive changed estimates.
- 4.Distinguish drafts, approvals, and verified external actions.
What carried forward
- A separate Second Brain is optional. Begin with relevant context and two current operating records.
- Uploading the PDF starts guided setup. Source access, saved instructions, and working routines still need verification.
- Clear proof creates a stopping condition. Any usage savings depend on the workflow and should be measured.
- The questions focused on preparation, source access, personal boundaries, and the economics of repeated use.
What to test next
- Use the build guide to create a first version from your existing responsibilities and goals.
- Try the fictional Hobson exercise separately, copying the specified prompt blocks.
- Change one input, save the records, and verify the result in a fresh conversation.
Take the work with you