The Operating Brain

A SaaS founder reviews half her team's pull requests because no junior ships at her bar.

An agency owner re-reads 80 client deliverables a week because the playbook lives in his head.

An ops lead spends Mondays reconstructing what got decided on Slack on Friday.

Different roles. Same leak. The bar, the playbook, the decisions: none of them are written down anywhere a new hire or an AI could read.

Decisions per day went up by an order of magnitude in the last eighteen months. AI made many of them. Most of those AI decisions are invisible. Nobody can pull a thread back to its source. Nobody can read why a draft got written this way, why a lead got disqualified, why a meeting got rescheduled. The audit trail evaporated the moment the work scaled.

Pre-AI, knowledge already leaked. It leaked through Slack threads nobody searched, through emails buried in PDF chains, through spreadsheets four people knew the schema for. Companies built on top of that leak anyway, because the volume was small enough to absorb.

Post-AI, the volume isn't small. Compounded over eighteen months, the company looks nothing like the one you started. The intelligence is everywhere and nowhere at once.

“Your org's intelligence sits idle across disconnected databases.”

An Operating Brain is the persistent system your company uses to capture decisions, codify expertise, and run rituals across every channel. It runs on three layers. Intelligence compounds what you know. Tools execute on what you know. Hygiene keeps the team aligned.

Intelligence has two surfaces. A vault holds unstructured knowledge: lessons, decisions, observations, blueprints. A structured database holds operational data: leads, accounts, campaigns, rolling metrics.

The vault is portable, AI-readable, and survives any vendor switch. Plain text with structured headers, version-controlled, indexed by topic and by date. The structured database is yours, hosted on infrastructure you control, queryable from any tool you bring in next quarter.

Underneath both surfaces sits an eight-layer stack. The bottom layers move slowly: your worldview, your soul, your objectives. The middle layers carry your lessons and your reference material. The top layers move daily: a synthesized briefing every morning, an operations queue for what needs action today.

The full eight-layer stack is covered in our companion roadmap, available below as a free download.

“Your company has memory now.”

The substrate is Claude. The runtime reads and writes against the tools your team already uses. Five capabilities matter at the leader level.

The connectors

Connectors wire the brain to the tools your team already uses: email, calendar, chat, project management, design files, the browser. The brain reads from each one and writes back when you give it permission. Compliance approves the scopes once. The brain inherits everything afterward, and your team stops copy-pasting between tools.

Read the strategic guide for executives

The codified expertise

Skills are reusable templates of how your company does specific work. A negotiator skill that knows your pricing and your past objections. A financial-model skill that follows your assumptions. A security-audit skill that checks against your compliance posture. The right skill surfaces for the task. Your expertise becomes shareable across every team without retraining.

Why Skills are the missing piece in most AI investments

The specialists

Subagents are isolated workers, each with its own tools and memory. One audits sales channels while another reviews code while another drafts copy. They run in parallel and report back. Work that used to be sequential becomes concurrent.

The rituals

Hooks are automatic triggers that fire at session start, session end, or after specific actions. Morning context arrives before the team logs in. Outcomes get pushed back into the vault when work ends. Isolated actions become orchestrated pipelines that nobody has to remember to run.

What's matured in the last six months

Context windows hold a year of email or a thousand-page deal room. Agents run in parallel, not in sequence. Files upload once and reuse across hundreds of requests. Every claim traces to its source. Agents wake on a schedule or on a webhook. Plug-and-play workflows install from an open marketplace. The substrate matured faster than most teams adopted it. Your competitive variable is the rate you adopt the next one.

Hygiene is the discipline that turns the substrate into a working system. Plans before code. Handoff documents at the end of every session. Sub-sessions over long ones. Role-specific skills assigned to roles. Predictable folders so anyone navigates the project on day one. None of it is glamorous. All of it compounds.

Hygiene is also the layer most companies skip. They buy the tools and forget the rituals. The output looks professional for a quarter, then drift sets in: contradictions stack up, the audit trail breaks, the AI starts hallucinating in confident voice. Layer 3 keeps the substrate honest.

“Twelve people ship like one senior.”

We use a customized fork of Hermes, the open-source AI runtime, as our Operating Brain substrate. We picked it over OpenClaw because Hermes runs a self-improvement loop in production. Every approved page, every flagged contradiction, every successful agent run feeds back into how the next session behaves. The runtime gets sharper while the team sleeps.

Both Hermes and OpenClaw are legitimate open-source choices. The reason this page leans Hermes is the loop. If your company's edge is the rate at which it learns, that is the variable to optimize.

Custom-built makes sense for mature engineering teams who want control of every layer. Slack-bots make sense for day-zero starters who want a single-channel proof of concept. The middle ground belongs to the open-source runtimes.

Read the OpenClaw versus Claude Code comparison

Five things change once an Operating Brain is in place. They compound.

1. A daily briefing every morning, curated across every channel.

   Email, calendar, chat, project tools, meeting transcripts. The brain synthesizes it into one document and pushes it to your phone before you finish coffee.

2. Decisions become searchable in seconds, not days.

   Every meeting outcome, every Slack thread that mattered, every approval the team granted. The audit trail is structured, the search is semantic, the answer arrives faster than the meeting that produced it.

3. AI work becomes auditable.

   Every action an agent took, every claim it made, every source it referenced. When something goes wrong, you trace it. When something goes right, you replicate it.

4. Onboarding takes days, not weeks.

   New hires read the vault and ship by the end of week one. The brain explains why decisions were made, who to ask about edge cases, what landed and what didn't. The senior's head stops being the bottleneck.

5. The founder takes two weeks off and the company keeps producing.

   Briefings keep arriving. Workflows keep running. Approvals queue for return. The brain holds the line.

“Your company already has a brain. It just lives in seven inboxes, a thousand Slack threads, and your top performer's head.”

Eight statements. Tick the ones that describe your company today. Each tick is a leak. The result panel tells you where to start.