A concept, explained
The Mental Model.
The brain for company management and agentic AI.
Robert Blust · Software Engineer & Architect
The hook
A company's knowledge lives everywhere — and nowhere.
Vision, strategy, processes, roles, KPIs, rules — scattered across heads, documents and tools.
Neither people nor AI can rely on it. What if all of it were a graph a machine could read?
The idea
One structured knowledge base — the brain of the company.
I call it the Mental Model: the facts as Markdown, structured by a meta-model.
Not new documentation but a machine-readable projection of the company's operating system — versioned like code, synced from the real systems.
The breadth
Everything that makes a company — in one place.
Roles & Teams who does what
Processes & Gates how work flows
Strategy & Objectives where we invest
KPIs what we measure
Rules what we enforce
Features what we build
Concepts what we define
Architecture decisions what we decided
Values what we believe
Versioned like code, machine-readable, ONE source of truth.
The graph
Everything cross-references — no dead links .
Person
Process
KPI
Rule
Feature
Role
Mental Model
Filenames are IDs the canonical name of every entry Exact-name references no markdown links that rot Validated on every change a contradiction surfaces immediately
An example
Governance a machine can reason over .
Owner Head of Product · gate sign-off Gate criterion Tech-Debt Impact Score (−3 … +3) Rule score > +2 → cleanup enters scope Kill criterion adoption < X% → deprecate
Rules, metrics, decisions
Rules, metrics and decisions — structured and owned.
Rule “Deployments follow dev → test → prod.” — Owner: Engineering.
KPI “Lead Time for Changes” — lower is better. — Owner: Engineering.
Architecture decision “JSONB only for contextual data.” — rejected: querying on it.
Every rule, metric and decision: named, justified, owned.
Two audiences, one source
The same model — for people and agents.
For people
Onboarding by role, decisions, alignment.
“Who owns the deployment?” — answered from one source.
For agents
A context layer plus MCP tools.
Agents act inside vision, strategy, process and rules.
One source — two audiences, no drift.
Why it matters
The right context makes AI far more reliable .
Without context, AI knows neither your decisions nor your rules — it guesses. The Mental Model shrinks that space and keeps agents inside the guardrails. Always with human oversight.
Takeaway
Explicit knowledge pays off twice — for people and for agents.
Structure is the lever. The cleaner the model, the better people lead and the more reliably agents act.