Knowledge & data

Practice knowledge base, light

How does the referral work again? What contact time does the new surface disinfectant need? Who covers for us over the holidays? In most practices the same experienced assistant answers questions like these – between two patients at the front desk.

AI knowledge system

01Intro

The light practice knowledge base makes the organisational processes and standards findable: prescription and referral paths, appointment types and durations, hygiene checklists, device instruction, holiday arrangements, the onboarding of new assistants. As a structured system the team looks things up in instead of asking around – and where AI answers the standard questions, with a source. Deliberately "light": the value lies in the queryable process knowledge, not in an automated full build. And the most important boundary is firmly drawn in: patient data doesn't belong in it – it stays in the practice-management system, where it belongs. The system holds how you work, not who you treat.

Up front: patient data stays out – that's not a matter for negotiation but the basic condition. The knowledge base holds processes and standards, the PMS holds the records. And "light" is deliberate: start small, one area first, make it queryable. A fully agentic system with autonomous routines is rarely needed in a practice – and only conceivable with a strict data-protection concept.

02Demo

simulation · fictional practice

The light practice knowledge base of the fictional “Dr Winter Practice” – on the left the order in which processes and standards live (patient data deliberately stays out), on the right three everyday moments with it. Click through and watch where it lights up in the tree.

// the knowledge vault
// what happens

The architecture behind it has two layers: at the bottom the vault – structured folders and notes that belong to the business. Above it the AI layer with documented rules: it reads only at first, then writes with approval, and sensitive areas stay locked. Calendar and tools dock on, routines run within clear limits.

// docked on Calendar · email Tools & automations VaultFolders · notesbelongs to the business AI layerreads firstwrites with approval ResultsAnswers · draftsRoutines Humanapprovesstays in charge // the rules (documented) sensitive areas locked · each routine allowed individually

The most honest proof of this model: this website came about exactly this way – I work in such a system myself every day, with AI as an employee inside it.

03How it works

  1. 01

    Take stock: where does your process knowledge live?

    In heads, on notes in the consulting room, in an old "practice manual" folder no one opens. Plus: which questions interrupt whom most often, and what does only the longest-serving assistant know?

  2. 02

    Build the structure, small and documented

    A clear ordering system: processes, standards, team organisation. With simple rules – and the firmly drawn-in line that patient data doesn't belong in it. That's the most important step.

  3. 03

    Fill one area first

    Usually the prescription/referral paths and the appointment types – the topics with the biggest crowd at the front desk. Existing material (manual, notes) moves in curated, not copied.

  4. 04

    Make it queryable

    The team – and especially new assistants – can ask the knowledge base and get answers with a source. That's the core of the light variant: look up instead of interrupt.

  5. 05

    Optional: write along with sign-off

    Once it holds, the AI can suggest changes to standards (a new hygiene rule, changed holiday dates) – but only as a draft, carried out after sign-off. A practice rarely needs more.

04Stages

For practices, Tier 1 is usually the goal, not the entry – queryable process knowledge already solves most of the problem.

Tier 1

Entry & core: the queryable knowledge base

Processes, standards and team organisation are structured and made queryable. New assistants look things up instead of interrupting the colleague at the desk. For many practices, that's exactly enough.

Tier 2

Extended: the AI writes along (with sign-off)

Changes to standards and processes are prepared as a draft and carried out after sign-off – the base stays current without anyone maintaining it by hand.

Tier 3

Only with a data-protection concept

A fully agentic system with autonomous routines is rarely needed in practices and only conceivable with a strict data-protection concept – deliberately not a standard offer.

05Technical implementation

  • The knowledge layer belongs to the practice: a Markdown vault (e.g. Obsidian), Notion or a comparable system – content without vendor lock-in.
  • The AI layer is an assistant with access to the vault (e.g. Claude Code as the agentic layer) – with documented rules on which areas it may read and (with sign-off) write.
  • Hard system boundary: patient data stays in the practice-management system; the knowledge base holds processes and standards only.
  • Tiered rights logic: reading first, writing only as a draft with sign-off – sensitive areas excluded.
  • Upkeep routines are set up alongside – but lean; the light variant lives on a maintained core, not on automation.
  • Typical content: prescription/referral paths, appointment types, hygiene and device standards, holiday/cover arrangements, onboarding of new assistants.

for context · This shows how something like this usually gets built – not how it has to be built for you. What actually fits is worked out in the project: around your existing setup, your team, and what stays easy to maintain.

06Related building blocks

See all levers for your practice