Lakoona

The living model of your organisation

Model how your organisation actually runs as one governed knowledge graph — then let Lakoona generate your policies, compliance evidence, risk register and staff training from it, and keep them in sync as it changes. Every answer grounded, every change reviewed and attested.

The idea

Documents are projections. Model the truth instead.

Underneath every policy, register and training package sit a small set of facts: who does what, what they use and produce, what can go wrong, what keeps it safe — and who owns it. Model those facts once, in a structure a machine can check, and every document becomes a view that can be generated, verified and regenerated when the facts change.

performsperformsprecedesconsumesproducesintroducescontrolsmitigatesownsRegistered NurseClinical LeadPrepare MedicationAdminister MedicationTwo-Person CheckMedicationAdministration RecordMedication ErrorROLEROLEFUNCTIONFUNCTIONCONTROLCONSUMABLEDATARISK
A miniature of the real thing. Every node is one of ten categories, every edge one of only thirty valid combinations — the system rejects anything else. Modelled once and machine-checked, this little graph becomes an answer, a compliance evidence trail, a risk-register row, a policy and a training module — the rest of this page runs on it.

Stop maintaining documents. Maintain the truth.

Knowledge graph

Your organisation, modelled and governed

Roles, functions, controls, records and risks — modelled as a constrained graph, not a wiki. Ten categories of things, seventeen relationship types, and only thirty valid ways to combine them, so the model stays clean and the system rejects structures that don’t make sense.

  • Built in conversation — upload a handbook or just describe a process
  • Ontology-validated when the AI proposes a change, and again at publish
  • Nothing goes live without review: authors can’t approve their own changes

A constrained model, not a wiki

10categories of things — roles, functions, controls, risks, records…
17relationship types
30valid combinations — everything else is rejected

Ask

Ask anything. Every answer shows its evidence.

Ask “how do we administer medication — and what keeps it safe?” in plain language. Lakoona answers from your published graph alone — the Registered Nurse prepares, then administers under the Two-Person Check, recording each dose in the Administration Record — with every claim a clickable citation.

  • Grounded entirely in your approved graph — never invented
  • Every citation machine-verified to resolve to a real node
  • Save any answer as a living, audit-grade document

Ask

“How do we administer medication — and what keeps it safe?”

The Registered Nurse first prepares the medication, then administers it under the Two-Person Check, recording each dose in the Administration Record.

Every underlined term is a live, machine-verified citation to a real graph node.

Compliance

From a 100-page standard to evidence you can defend

Upload the standard and Lakoona breaks it into individual obligations with page-and-clause provenance, then evaluates each against your actual operations. Evidence links to the graph itself — the Two-Person Check, the nurse who administers, the record each dose produces — so it can never point at something that doesn’t exist.

  • Every obligation graded against your real roles, controls and processes
  • Evidence is FK-enforced links to real graph nodes, not attached PDFs
  • A gap is the exact missing relationship — “no control mitigates Medication Error”

From standard to evidence

414obligations extracted from the NDIS Practice Standards
82sections, each with page & clause provenance
120obligations from the Australian Privacy Principles

✗ gap — no control mitigates Medication Error

A gap is stored as the exact missing relationship, not prose.

Internal extraction benchmarks, 2026.

Risk

A register you don’t maintain

Functions introduce risks; controls mitigate them; roles own controls. Those three facts from your graph are a register row — Medication Error, introduced by administration, mitigated by the Two-Person Check, owned by the Clinical Lead — so exposure, coverage and accountability are computed, never re-typed into a second system.

  • Set only likelihood, consequence and control effectiveness — the rest is derived
  • Strengthen a control and residual risk updates everywhere
  • Uncontrolled risks and unowned controls surface themselves, with guided fixes

A register row, computed

  • Administer MedicationintroducesMedication Error
  • Two-Person CheckmitigatesMedication Error
  • Clinical LeadownsTwo-Person Check

Exposure, coverage and accountability fall out of the graph — no second risk system.

The paperwork stays in sync with reality

Because policies and training are views of the graph, a change to how you work updates them — you don’t.

Documents

Policies that revise themselves

Your Medication Management procedure is authored with the assistant, section by section, every claim cited to the graph. Tighten the Two-Person Check and the procedure is found automatically, judged for material change, and revised — edited to reflect exactly what changed, with the previous version kept in history.

  • Revision, not regeneration — the document keeps its shape
  • Cosmetic changes cost nothing and notify no one
  • Every part of your organisation gets a system-managed overview

Training

Training that’s never out of date

Modules and quizzes come from the same graph your policies do — “what must happen before you administer medication?” is grounded in the Two-Person Check itself. The Registered Nurse role requires the module, so every nurse and whoever covers for them is assigned it automatically.

  • Assigned by your org structure, hierarchies and stand-ins included
  • Change a tested control and only affected staff are re-certified
  • A re-worded intro never expires anyone’s certification

The flywheel

Change once. Everything follows.

After a near-miss, you strengthen the Two-Person Check on medication administration. One conversation with the assistant —

  1. The graph is updated — reviewed, approved by the Clinical Lead who owns the control, attested.
  2. The Medication Management procedure revises itself — version history kept.
  3. The medication module updates only if what it tests changed — and only staff who administer medication are re-certified.
  4. The residual rating on Medication Error recalculates.
  5. Compliance evaluations citing the check are flagged stale, ready to re-run.
  6. The next person who asks “how do we administer medication?” gets the new answer — with citations.

Not five tools. One model with five projections.

Built for the audit

Audit-grade by construction

For organisations whose processes are examined — NDIS providers, aged care, schools — governance isn’t a feature bolted on. It’s how the graph works.

Nothing is ever overwritten
Every version of every fact is retained — who changed it, when, in which revision, and what it looked like before.
Attestation is first-class
Structural changes require attestation by the accountable owner; you can see, at any time, what fraction of your model a named owner has reviewed.
Four-eyes by default
Authors can’t approve their own changes. Every affected domain needs its owner’s sign-off before anything goes live.
Isolated by construction
Each organisation’s data is isolated at the database itself with row-level security — not just in application code.
Grounded, cited, checked
Answers and generated documents are built from your published graph; every citation is machine-verified, and fabricated references cannot render.
Pinned to revisions
Answers, documents, evaluations and training are all pinned to graph revisions — you can always say what the model looked like when a decision was made.

See Lakoona on your own operations

Bring your staff handbook and one standard — we'll model the handbook with your quality lead and show you the first compliance evaluation. Send a note and we'll set up a demo, or email us at enquiries@lakoona.com.au.