Amy
Axelrod

  • Sum of All Parts
  • All There
  • The Garden of Forking Pawns

All There: A Better Briefings System

This case study examines executive briefings: one person carrying a never-ending calendar of meetings and the briefings they require. When a system can assist with research, filing, and reminders, every meeting can get the same care, and preparation scales without a cost to quality.

Please email amydaxelrod [at] gmail [dot] com
for resume requests.

LinkedIn

All There

Click to enlarge
1 / 5·

Notes

Independent concept based on experience as a chief of staff. Illustrative data. No confidential information shared.

Reasoning

A great meeting is not the hard part. The hard part is everything that comes after. The follow-up, the notes, and remembering who someone was and what they cared about the next time their name comes up can easily get lost. When it is one person doing this on top of everything else, most of it slips, not from lack of effort, but because there is no system to catch it.

This results in promising connections fading between meetings, and every new meeting starts from zero.

This case study argues that scale should not come at the cost of quality.

References

In The Myth of Sisyphus (1942), Albert Camus uses the figure condemned to push a boulder uphill forever as an image of repetitive labor without end. The slide borrows the image for a calendar that resets every week. Must we imagine Sisyphus as happy?

Albert Camus, The Myth of Sisyphus, trans. Justin O'Brien (New York: Vintage International, 1991).

Keywords: executive support · stakeholder management · relationship management · root-cause analysis

Independent concept based on experience as a chief of staff. Illustrative data. No confidential information shared.

Reasoning

A first meeting only goes well if the principal walks in prepared, and a relationship only continues if something keeps the history. That is the part that compounds.

This is not about doing less. The principal still shows up as themselves, and the chief of staff still makes sure they put their best foot forward.

References

First-principles thinking means breaking a problem down to what has to be true, then building back up from there instead of copying how it has always been done. Applied here, it separates what only the principal and chief of staff can do from what a system can carry.

Shane Parrish, "First Principles: The Building Blocks of True Knowledge," Farnam Street, https://fs.blog/first-principles/.

The research on the role points the same way. Antoine Buteau describes a chief of staff's real work as judgment, not summarization: signal versus noise, a one-off versus a pattern, and an operating memory where past decisions stay retrievable.

Antoine Buteau, "The Chief of Staff Operating Model, 4: Context Flow," https://www.antoinebuteau.com/the-chief-of-staff-operating-model-series-4-context-flow-making-the-company-legible-to-leadership.md.

Keywords: first-principles thinking · prioritization · capacity planning · calendar management

Independent concept based on experience as a chief of staff. Illustrative data. No confidential information shared.

Reasoning

The layer. Everything runs from one AI project that holds the context: the principal's profile, goals for each relationship, trust rules, voice and tone, and every briefing and note so far.

Outlook, Google Drive, a tracker spreadsheet, and Trello are connected to it. Zapier catches the exact moments a person shouldn't have to watch for: an approved briefing saved, a meeting summary arriving, the end of the month. Meetings enter the system by keyword: the chief of staff adds [CoS] to an invite's title, and the morning run and Zapier pull only what carries it, so personal and unrelated events stay out.

This slide is detailed on purpose. Like a laptop, the system underneath is complex so the result on top can be simple: one page the principal can read in a minute. Every meeting goes through the same steps, so a last-minute meeting gets the same care as one booked weeks out.

The judgment stays with the chief of staff: which briefing a meeting needs, which sources to trust, what stays private, and what goes out in the principal's voice. The system drafts and remembers. Nothing goes out until the chief of staff approves it.

Briefing Types

  1. Bespoke. A priority person, with an institution behind them. The full structure.
  2. Standard. Lower stakes, shorter, more general.
  3. Multi-person. Three or more people at once, for example a lunch during a trip. One glance card for the group, then a short block per person and how they connect.
  4. Board. Straightforward, no talking points.

A briefing is almost never longer than one page unless asked. Few but purposeful words.

The Research Agent

The research agent is configured, not trained, and the rules are specific:

  1. Identity check first. A person is matched on at least two facts, such as employer, role, or location. Two people with the same name are never merged.
  2. Sources are ranked: their own site or organization first, then established press, then interviews and talks.
  3. Every claim is dated, and older claims are flagged.
  4. When sources disagree, both are shown.
  5. “Not enough reliable information” is a valid answer. The agent does not fill gaps.
  6. Each fact carries a confidence label: confirmed, likely, or unclear.
  7. Read only, with a step limit. Nothing is saved or sent without review.

For people met before, a “what's changed” agent reports only what is new since the last meeting, so the principal starts with what changed instead of repeating last time.

When the Day Changes

  1. Last-minute meeting. Add the person by hand, or ask the project for a short briefing.
  2. A busy day. The tracker is the queue, and reminders repeat three days before, one day before, and the morning of.
  3. Several meetings at once. One tracker row per meeting, and choosing the briefing type is the triage. For example, seven important meetings in three days is still manageable.
  4. A forgotten follow-up. An email reminder the next morning if the “follow-up sent” box is empty.

Risks and Mitigations

RiskMitigation already in the design
Researching the wrong personIdentity check on two facts, the chief of staff picks the sources
Outdated informationDated claims, confidence labels, the what's-changed agent
Confidential details in a briefingTrust rules (private, escalate, pattern only), nothing sent without review
Over-trusting draftsThe chief of staff approves every briefing and follow-up
A missed inviteInvites without the [CoS] keyword are ignored, so the morning digest lists new events for a quick check
Cancelled or moved meetingsCancellations are set by hand, the next morning run catches moves
No meeting summaryThe host decides on recording, a voice memo after the meeting is the fallback
The principal doesn't read itA glance card first, readable in ten seconds

Considered and Rejected

  1. Auto-sending follow-ups. The relationship is the human work. The AI drafts in the principal's voice, and the chief of staff sends.
  2. An app the principal has to use. The principal should never have to touch the system. They get one page.
  3. Recording without consent. Recording is the host's decision, and consent rules vary by state and organization.

Time Saved

Illustrative, based on my own experience preparing 30+ briefings. A bespoke briefing used to take two to three hours: at least an hour of research, then copying notes together, deciding what to say, editing, and editing again in the principal's voice so they could read it quickly.

TaskBeforeWith the system
Bespoke briefing2–3 hrs30–45 min (choose sources, review, edit)
Standard briefingAbout 1 hrAbout 15 min
Follow-up email20–30 min5–10 min (review and send)
Logging a new invite and filing notesAbout 20 min per meetingAbout 2 min (a quick check)
“Who had the idea for X?”15–30 min of searchingUnder a minute
Monthly reviewAbout 1 hrAbout 10 min
A week of five meetings (two bespoke, three standard)About 10–13 hrsAbout 2–3 hrs

Build

This is a one-person system, so the build is short: about two weeks, alongside the job.

  1. The AI project, context, and skills (2–3 days).
  2. The tracker spreadsheet (1 day).
  3. Three Zaps: the approved briefing, the meeting summary, the monthly Trello comment (1–2 days).
  4. The research and what's-changed agents, tested on public figures (2–3 days).

The same template could be shared across an executive assistant team or an office of the chief executive, with each person keeping their own project.

References

Paul Rand, the designer of the IBM and UPS identities, wrote that simple designs are “the products of circuitous mental purposes. Simplicity is difficult to achieve, yet worth the effort.” The slide is the circuitous part. The briefing is the simple part.

Paul Rand, Design, Form, and Chaos (New Haven: Yale University Press, 1993).

The design separates workflows (fixed, predictable steps, like the Zaps) from agents (open-ended research), following Anthropic's guidance to use the simplest pattern that works and add autonomy only where it helps. That post now carries a note that the tooling has changed. Today, the research and what's-changed agents could run on Claude Managed Agents (in beta), Anthropic's hosted agent harness, built for long-running and scheduled work in a sandbox, with credentials kept out of the agent's reach. That last point matters in a role where trust is the main constraint.

Anthropic, “Building Effective Agents,” December 2024, https://www.anthropic.com/engineering/building-effective-agents.

Anthropic, “Claude Managed Agents Overview,” Claude Platform Docs, https://platform.claude.com/docs/en/managed-agents/overview.

Lance Martin, Gabe Cemaj, and Michael Cohen, “Scaling Managed Agents: Decoupling the Brain from the Hands,” Anthropic, April 8, 2026, https://www.anthropic.com/research/managed-agents.

OpenAI's guide reaches the same conclusions: use an agent only where judgment or unstructured information makes fixed rules hard to maintain, layer several guardrails rather than relying on one, and hand off to a person before any high-risk or irreversible action. Here, that hand-off is the chief of staff's approval before anything is sent.

OpenAI, “A Practical Guide to Building Agents,” https://openai.com/business/guides-and-resources/a-practical-guide-to-building-ai-agents/.

Keywords: workflow automation · AI agents · Zapier · prompt design · research and synthesis · human-in-the-loop · process design

Independent concept based on experience as a chief of staff. Illustrative data. No confidential information shared.

Reasoning

I think of the moment in The Devil Wears Prada when, at the gala, Andy and Emily deliver every name to Miranda exactly when it's needed. Miranda, and people on her level, such as heads of state and executives, must show up in every room as themselves, despite an endless schedule. Their briefings are built for any setting, on the assumption they might only be glanced at. The briefing has to work even when the principal doesn't sit down to read it.

The briefing on top is the plain thing, ready to use. Everything underneath it, the research, the history, the context, is what makes it precise.

Over time, the context compounds. A quarterly review asks, “These are everyone you met. Who do you want to keep up with?” Enough accumulates that the briefings could feed a newsletter. And a question like “who did we talk to three months ago who had the idea for X?” is answered in seconds, with full context.

Measurement Plan

  • Log one baseline month: time per briefing, follow-ups sent, and how many within 48 hours.
  • Run one month of real meetings with the system, on the same principal and calendar, then adjust with the principal's feedback on the glance card.

Measure

  • Meetings briefed
  • Follow-ups sent within 48 hours
  • Time per briefing
  • Relationships kept warm each quarter
  • The principal's own read: did they feel prepared?

Limits

  • It is only as good as the notes that come back from each meeting.
  • Review time is the ceiling. The chief of staff still approves everything.
  • Trust rules have to be agreed with the principal first.
  • Public research depends on what is public.

What was built and what is proposed: I prepared 30+ briefings and set up workflows and automations in Trello, Outlook, Zapier, and Calendly in the role. The full system on these slides brings that practice together into one design.

References

McKinsey's guide to the role puts it plainly: “Your job is not to make yourself look good but to make your principal look good.”

McKinsey & Company, “Seeing Around Corners: How to Excel as a Chief of Staff,” https://www.mckinsey.com/capabilities/strategy-and-corporate-finance/our-insights/seeing-around-corners-how-to-excel-as-a-chief-of-staff.

Keywords: executive briefings · stakeholder management · relationship building · metrics · baseline

Independent concept based on experience as a chief of staff. Illustrative data. No confidential information shared.

Reasoning

The judgment stayed with me: what to look at, what mattered, what to keep. I dictated each step of the system and every paragraph. AI widened what I could see, explained the options, and rendered slides 2 to 5 through dialogue, but the choices were mine.

References

W. Edwards Deming (1900–1993) was the statistician whose work shaped modern quality management. He held that most problems come from the system, not the individual worker, an idea often paraphrased as “a bad system will beat a good person every time.” The Discover row starts from the same place: relationships fade not from lack of effort, but because there is no system to catch them.

“W. Edwards Deming,” Wikiquote, https://en.wikiquote.org/wiki/W._Edwards_Deming. (Paraphrase; no confirmed original source for the exact wording.)

Keywords: AI fluency · responsible AI use · human-in-the-loop

·

Notes

Click the slide to zoom