M7 · HOW I RUN MY DAY · OP1

The Chief-of-Staff Playbook

Build the chief of staff. Don’t hire one. Email triage, transcript handling, brain capture, daily research.

10 SLIDES · ~12 MIN · PLAYBOOK OVERVIEW
Key Takeaways
  • Build the chief of staff. Don't hire one.
  • 200 — emails a day across personal and BuildClub inboxes (BuildClub — the AI operating system Stephen Forte runs his firm on).
  • 170 wpm — dictation speed I sustain with Spokenly — faster than typing or talking.
  • ~10 — meetings a week, each one a transcript that becomes a working doc.
  • Pick one inbox. Build a morning-brief agent against it this week. Approval, not autonomy.
MODULE 7 · OPERATOR OP1

The Chief-of-Staff Playbook

You don't hire a chief of staff. You build one.

BuildClub Academy
01 · WHY THIS MATTERS

Three numbers that explain why this matters.

200
emails a day across personal and BuildClub inboxes (BuildClub — the AI operating system Stephen Forte runs his firm on).
Personal volume
170 wpm
dictation speed I sustain with Spokenly — faster than typing or talking.
Spokenly
~10
meetings a week, each one a transcript that becomes a working doc.
Fireflies + Zep + Notion stack
02 · THE THESIS

Build the chief of staff.
Don't hire one.

The Thesis
03 · MORNING EMAIL AGENT

The morning email agent — four steps before I'm awake.

01
Pull overnight inbox
Agent sweeps personal + BuildClub inboxes via Gmail API on a 5am cron. No human in the loop.
02
Classify and prioritize
Each thread tagged: action / FYI / noise / commitment. Anthropic model with my voice + priorities in the system prompt.
03
Draft responses
Action items get full draft replies as Gmail drafts. Tone matches mine because the agent has read every email I've ever sent.
04
Daily brief at 6am
One email lands in my inbox: today's commitments, drafted replies awaiting approval, what slipped, what is at risk.
04 · INSTITUTIONAL MEMORY

The brain: Zep + Mem0 share the memory across every agent — and let me change AI tools without losing it.

Takeaways
  • 1 Zep stores the graph of entities, decisions, and relationships across meetings, email, and docs. Not a vector dump — a graph the agents query before they reply.
  • 2 Mem0 records the memory my AI tools form during a session — what I asked, what I corrected, what I prefer — and makes it portable across tools.
  • 3 Portability is the point. When I switch from one AI tool to another next year, my Mem0 memory follows me. The agents wake up already knowing who I am and how I work.
  • 4 When I tell the chief-of-staff agent something, the build agent and security agent know it too — through Zep. When I switch tools, the new tool knows it too — through Mem0.
05 · TRANSCRIPT WORKFLOW

Every meeting becomes a working doc — automatically.

01
Fireflies captures
Auto-joins every Zoom/Meet. Records, transcribes, speaker-labels. No manual setup.
02
Classify and route
Agent reads the transcript: client, project, decision type. Routes to the right Notion page.
03
Extract decisions
Action items, commitments, open questions — pulled out as structured fields, not buried in prose.
04
Feed Zep
Decisions and entities written to the graph. Future queries surface this meeting automatically.
05
Draft the follow-up email
Within 10 minutes of the call ending, a draft recap email is sitting in my Gmail drafts waiting for approval.
06 · FILE INFRASTRUCTURE

The five-surface file infrastructure.

1
2
3
4
5
1
Notion is the human-facing surface
Every project, every client, every meeting has a Notion page. Humans read here. Agents write here.
2
Google Drive is the human knowledge layer
Documents I share with my team, clients, and collaborators. Where human knowledge lives, gets reviewed, and gets handed off. The everyday read-write surface for the people on my team.
3
GitHub is the agent surface
Prompts, scripts, configs, audit logs. Versioned. Diffable. Reviewable. Agents read and write here.
4
S3 is agentic storage
The materials and knowledge my agents need to do their work — raw transcripts, raw analyses, structured corpora, model artifacts. Not the docs I read day to day; the corpus the agents read.
5
Zep + Mem0 sit on top
The brain that lets agents reason across Notion, Drive, GitHub, and S3 without re-reading everything every time — and the portable session memory that follows me across AI tools.
07 · DAILY RESEARCH

The daily research engine runs on its own schedule.

Takeaways
  • 1 Every morning a Perplexity Computer agent runs a 9-topic AI news sweep, dedupes against yesterday, drafts a Notion brief.
  • 2 The same pattern powers the daily podcast: research, script, voice, mix, publish — all overnight.
  • 3 I edit, not author. The engine does the cold start; I do the taste layer.
08 · THE DICTATION HABIT

The habit that compounds: dictate, don't type.

1
Spokenly runs everywhere
Hold a hotkey, talk, release. Output is clean, punctuated, formatted text — in any app, any field, any model. Macroses, custom vocab, and per-app profiles compound the speedup over time.
2
170 words per minute sustained
Roughly 3× faster than typing for me. Compounds across emails, prompts, docs, and code comments — every input I make all day. Multiply that across thousands of inputs and the leverage is the day, not the keystroke.
3
It changed how I prompt
I prompt the way I would brief a chief of staff in person — full sentences, full context. The model gets a better question, gives a better answer. Voice is the bandwidth match for how a CEO actually thinks.
CLOSING

Five things to do Monday.

Monday Morning
  1. 1
    Pick one inbox. Build a morning-brief agent against it this week. Approval, not autonomy.
  2. 2
    Turn on automatic meeting transcription for every call. No exceptions.
  3. 3
    Choose one memory layer (Zep, Mem0, or a Notion DB). Feed transcripts and decisions into it.
  4. 4
    Install dictation everywhere. Stop typing prompts. Brief the model the way you brief a human.
  5. 5
    Name the four file surfaces in your org: human-facing, agent-facing, cold store, brain. Anything that doesn't fit one of those four moves.
OP1 · Questions CEOs Ask

Frequently Asked Questions

What is the core idea of The Chief-of-Staff Playbook?
Build the chief of staff. Don't hire one.
What does the data say a CEO should pay attention to?
200 emails a day across personal and BuildClub inboxes (BuildClub — the AI operating system Stephen Forte runs his firm on). 170 wpm dictation speed I sustain with Spokenly — faster than typing or talking.
What should a CEO do Monday morning after reading The Chief-of-Staff Playbook?
Start here: Pick one inbox. Build a morning-brief agent against it this week. Approval, not autonomy; Turn on automatic meeting transcription for every call. No exceptions; Choose one memory layer (Zep, Mem0, or a Notion DB). Feed transcripts and decisions into it.
What are the steps in The Chief-of-Staff Playbook?
1) Notion is the human-facing surface; 2) Google Drive is the human knowledge layer; 3) GitHub is the agent surface; 4) S3 is agentic storage; 5) Zep + Mem0 sit on top.
Where do the claims in The Chief-of-Staff Playbook come from?
The playbook cites YPO Technology Network AI Daily Brief S1E57 — Humans Behind the Automation; BuildClub agent pipeline (Gmail + Anthropic + Apps Script); Zep — temporal knowledge graph for agents; Mem0 — portable memory for AI tools.
OP1 · Sources

All Sources in This Playbook

00 / 10