← The KlyHub Journal
ArchitectureMay 13, 2026 · KlyHub Team

The 5-layer ontology

The 5-layer ontology

When we set out to build KlyHub, the first design question was: what is the shape of a company that an AI should read?

A wiki is too flat. A graph database is too academic. A document tree is too arbitrary. We needed something that founders we admire would recognize as "how I already think about my company" and that an AI client could navigate in under three tool calls.

We landed on five layers. Each is independently useful. Each composes with the others. Each maps to a part of the founder's mental model.

Layer 1: Core

The smallest layer. The one you defend at the elevator pitch. Mission, vision, the "what do you do?" you say at parties. The founders' priors that don't change much quarter to quarter.

In KlyHub, Core entities include:

  • Company definition — the canonical pitch
  • Founders & senior team — who's responsible for what
  • Mission / vision / values — what you optimize for when in doubt
  • Constraints — what you've decided is non-negotiable (e.g. "no enterprise sales motion until ARR > $1M")

When an AI client opens a fresh conversation, it pulls Core first. Every downstream tool call gets to assume "we know who this company is."

Layer 2: Market

Who do you serve, who do you compete with, and how do you win? This is the layer where most pitch decks live. KlyHub turns it into a structured surface an AI can query:

  • Segments — your primary, secondary, anti-segments
  • ICPs — concrete personas inside each segment with discriminators
  • Competitors — direct, adjacent, indirect, with positioning notes
  • Substitutes — what customers do instead of buying you
  • Wedges — the angles you use to enter each segment

When you ask Claude "should we pursue the dev-tools wedge in Q3?", it has the Market layer to ground the answer. No more hallucinated competitor analysis.

Layer 3: Motion[]

The plural is important. Most companies run multiple go-to-market motions in parallel — and most knowledge-base products quietly force you into one.

KlyHub's Motion layer is an array of distinct, named motions. Each Motion has:

  • Type — sales-led, product-led, community-led, partnership-led, ads-led, founder-led
  • Owner — who runs it
  • Current state — exploring / scaling / sunsetting
  • Cohort — what kind of customer it produces
  • Metrics — what success looks like per motion

This is the layer that changes most. New experiments enter. Old motions get sunset. Your AI sees the full portfolio every conversation, not a frozen snapshot from your last all-hands.

Layer 4: Operations

How the work actually gets done. This is the boring-but-load-bearing layer that most knowledge bases skip entirely.

  • Rituals — weekly standups, monthly business reviews, quarterly planning
  • Playbooks — onboarding, hiring, incident response, customer success
  • Finance posture — runway, burn, gross margin targets, fundraising stance
  • Hiring plan — open roles, sequencing
  • Vendors & tools — what you pay for, why
  • Compliance posture — SOC2, GDPR, HIPAA status

When you ask Claude "what does our onboarding look like?", it doesn't hallucinate. It reads the playbook.

Layer 5: Memory

What happened, what you tried, what you learned. The corporate journal an AI can consult.

  • Decisions made — ADR-style with context, rationale, alternatives
  • Experiments run — hypothesis, result, learning
  • Incidents — postmortems with corrective actions
  • Customer signals — quotes, complaints, expansion conversations
  • Strategic shifts — when did the plan change, why

Memory is what lets the AI say "you tried this six months ago — here's what happened. Want a different angle this time?" Instead of generic advice, you get advice grounded in your own history.

Why five layers?

We tested three, four, six, and eight. Five is the magic number for these reasons:

  • Discoverable — a founder can hold all five in working memory.
  • Composable — most useful queries touch 2–3 layers (e.g. "what motion serves this segment best?" = Market + Motion).
  • Stable — the layer set hasn't needed revision in the six months we've been using it internally with design partners.
  • AI-friendly — a single MCP tool call can list available layers; another can fetch any layer's contents; the AI has a navigation pattern it figures out in under three exchanges.

What the AI sees

When Claude (or ChatGPT, or Cursor) opens a KlyHub-connected conversation, the implicit toolset includes:

list_layers()                    -> the 5 layer names
get_layer(layer: string)         -> structured contents
search_entities(query: string)   -> cross-layer search
get_entity(id: string)           -> full entity with relations

The AI plans its tool calls. The user doesn't have to. After ten or so conversations, the AI has learned your information shape and starts asking the right questions of the data automatically.

What this isn't

It's not a CRM. It's not a project management tool. It's not a wiki. Those products serve operational workflows; KlyHub serves AI-readable strategic context. There's no overlap with where you track deals, run sprints, or write specs — those tools exist and do their jobs well.

KlyHub is the thing that sits alongside your CRM, your PM tool, your wiki — and gives your AI assistants a coherent picture of the company they're helping you run.

Try it yourself

The fastest way to internalize the 5-layer ontology is to take KlyHub's four-phase intake. Twenty minutes of guided questions and you'll have a non-trivial Core + Market + first Motion + initial Operations + a few Memory entries. From there, you ladder up over weeks.

Start your trial — your company shape will be ready before your morning coffee.

The 5-layer ontology · KlyHub