Concept one-pager

David's BOS
Hermes for the CEO

A CEO-facing operating system. It is a CRM, but its accounts are not customers. They are divisions, initiatives and decisions. It exists for one reason: so the person at the top is never blindsided, and never has to leak confidential data to find out what is going on.

accounts = divisions one permissioned source of truth council-verified signals model-agnostic core

The problem

Information arrives late and filtered. Every update passes through staff first, so by the time it reaches him it has been smoothed.
Divisions are siloed. Studios, Parks and Consumer Products each hold a piece of the same picture. Nobody holds all of it.
Misalignments surface after they cost money. He learns about the collision when two teams have already spent against each other.
Staying informed means chasing people. Today the loop is his own: ask, wait, react to a crisis, ask again.

The requirement is narrow and hard: current, unfiltered, cross-division visibility without exposing confidential material to anyone who should not see it.

The shape of the answer

Take the architecture of a personal operating system and re-point it at a company. The plumbing is the same. What changes is what flows through it, who may see it, and what has the right to interrupt the CEO.

Ingest everything

Slack and Teams, email, meeting transcripts, and the AI chats employees already use. The layer removes manual chasing, because nobody has to forward anything.

Remember by division

Each division keeps its own memory. Those memories unify into one graph, and the graph is permissioned, so unification does not mean exposure.

Filter before escalating

A council of division agents votes on whether a conflict is real. Only what clears the threshold reaches him. The threshold is what keeps the channel trustworthy.

Architecture

LAYER 1 · INGESTION Slack and Teams channels, threads Email and calendar commitments, invites Meeting transcripts decisions, owners Employee AI chats the work nobody files LAYER 2 · ORG MEMORY, PERMISSIONED Org memory graph per-division memory, unified into one source of truth with access rules attached LAYER 3 · DIVISION AGENTS, ALWAYS ON Studios initiatives, risks Parks initiatives, risks Consumer Products initiatives, risks Direct to Consumer initiatives, risks International initiatives, risks LAYER 4 · ALIGNMENT COUNCIL · FILTER, NOT MEGAPHONE One agent per division votes: is this conflict real? a detected conflict escalates to the CEO only after it clears the consensus threshold threshold met LAYER 5 · CEO CADENCE AND CLOSE THE LOOP 07:00 briefing only what is relevant Pre-meeting briefs who decided what, when Route in one click both leads, same message Resolution log closes the loop, visibly DISCRETION LAYER · every layer above is filtered by it
flow of signals discretion layer permissioned memory the council sits between detection and the CEO on purpose

Same plumbing, re-pointed

ComponentBOS equivalent for DavidPain it kills
WhatsApp bridge Ingestion layer: Slack and Teams, email, meeting transcripts, and employees' AI chats Manual chasing
Shared memory Org memory graph: per-division memory, unified into one permissioned source of truth Fragmented knowledge
Agent council with consensus threshold Alignment Council: one agent per division votes on whether a detected conflict is real before escalating it Late discovery, and notification overload
Agent swarm Division agents: each summarizes its division's initiatives, risks and commitments continuously Staff-filtered information
Scheduler with briefings and alerts CEO cadence: 07:00 briefing, pre-meeting briefs, weekly board prep, KPI threshold alerts Six-week-old decks
Web search External radar: press, analysts, social and competitor monitoring Hearing it from a journalist
Security hardening Discretion layer: role-based access, relevance routing, audit trail, no raw confidential data exposed Leak risk, and the trust gap behind it
Primary model Model-agnostic core: runs on whatever models each division already uses Vendor lock-in, and the friction from Legal and Security

What David actually manages

Divisions

Health score, active initiatives, owner.

Initiatives

Status, dependencies on other divisions, last real update.

Decisions

Who decided, what information they used, which divisions are affected.

Signals

A detected conflict or risk, moving through a pipeline. Every stage is visible, including the ones the council rejected.

People

Accountability map: who owns what, and who to ping when it slips.

Detected→ Council verified→ Routed→ Acknowledged→ Resolved

His daily loop

  1. Opens the morning briefing, which covers only what is relevant to him.
  2. Asks follow-up questions in chat, with answers pulled from real cross-division context instead of a forwarded summary.
  3. Gets council-verified misalignment flags before either team has to escalate them.
  4. Routes a flag to both division leads in one click, so both see the same message at the same time.
  5. Watches the resolution log close the loop, which is also what keeps the channel honest over time.

What to measure

01

Time from signal to CEO awareness.

02

Number of coordination failures avoided.

03

Precision of alerts: share rated relevant, not merely opened.

04

Meetings replaced, not meetings attended.

05

Adoption rate across divisions. This one decides whether the other four mean anything.