Skip to main content
CareLedger

CareLedger Assist

Ask your records.
Safely.

A safety-aware assistant on top of your live participant record. Ask in plain English, get a verified card from the same set your team already uses in the MAR. It is read-only and draft-only by design.

Start a free trial Read the AI safety brief

The promise

Three rules. Everywhere the assistant appears.

Read-only and draft-only

The assistant looks up, summarises, and drafts. It never signs a MAR cell, gives a medication, or changes a chart on its own. Every clinical write still flows through the audited workflow your auditor already approved.

Verified components, not freeform text

Answers render as the same cards your team uses in the MAR (med due, vitals trend, shift summary, draft message). The model picks which card; CareLedger supplies the data, the validation, and the rules.

Australian, audited, opt-out at any level

Records never leave the Sydney and Melbourne ISO 27001 region. Every prompt, tool call, returned field, and rendered card is logged and replayable. Any organisation, role, or participant can switch the assistant off.

What you can ask

Every question renders a verified card.

Look something up

  • "Who is due for meds in the next hour?"
  • "Did Mr Chen take his warfarin this morning?"
  • "Who is on shift tonight?"
  • "Anyone with a fever today?"
  • "Show me Lila's continence chart for this week."

Summarise

  • "What changed since yesterday?"
  • "Give me the handoff for the morning shift."
  • "Any incidents in the last 24 hours?"
  • "Summarise Marcus's vitals trend this week."

Draft (you press send)

  • "Tell Mr Garcia's daughter what happened today."
  • "Draft a shift-handoff note for the evening team."
  • "Write a progress note from these observations."

By voice

Coming soon
  • "What's on this afternoon's roster?"
  • "Mark Jordan's 10:30 metformin as given." (opens the MAR to confirm)
  • "Read me Priya's care notes."

One daily ritual

Morning handoff, two ways.

Before

45 min

  • Read every progress note from the night shift.
  • Cross-check the MAR for missed or held doses.
  • Scan vitals for anything out of range.
  • Re-type it all into the morning huddle email.

After

12 min

  • Ask “what changed since yesterday?”
  • Read one shift-handoff card: new orders, missed doses, vitals out of range.
  • Every row links back to the source entry in the MAR.
  • Confirm in the MAR before acting. The card cannot sign for you.

Under the hood, in plain English

How the assistant chooses a card.

  1. 1

    You ask. Your plain-English prompt goes to the model along with a fixed list of read-only tools it is allowed to call. Nothing else.

  2. 2

    It picks a tool. The model chooses one of the allow-listed tools and its arguments. CareLedger runs the tool against your live record and validates the result.

  3. 3

    It picks a card. The model selects which verified component to render. The data and validation are CareLedger’s, not the model’s; the model never writes the clinical values.

  4. 4

    You act. You read the card and act inside the audited MAR. Every step above is written to the audit log with the AI receipt below.

The AI receipt

Every card shows its working.

Each card the assistant renders carries a receipt: the prompt, the tool it called, the fields returned, and the card chosen. It answers “how do I know it didn’t make this up?” structurally, on every card.

AI receipt
prompt
"Who is due before lunch?"
tool
returned
4 entries (id, participantId, drug, dose, route, dueAt)
card
MedDueCard (sorted by dueAt)
logged
2026-06-20T10:14:33 AEST · user d.okafor · shift AM

See it on your own anonymised sample.

A 20-minute walkthrough: real prompts, real cards, and the AI receipt behind each one.

Book a demo Read the AI safety brief